lungcancerdetectionusingimagesegmentationbymeansof...
TRANSCRIPT
Research ArticleLung Cancer Detection Using Image Segmentation by means ofVarious Evolutionary Algorithms
K Senthil Kumar 1 K Venkatalakshmi 2 and K Karthikeyan 3
1Assistant Professor Department of Electrical and Electronics Engineering University College of Engineering Arni India2Assistant Professor Department of Electronics and Communication Engineering University College of Engineering TindivanamTindivanam India3Teaching Fellow Department of Electronics and Communication Engineering University College of Engineering Arni India
Correspondence should be addressed to K Senthil Kumar kslksenthilgmailcom
Received 30 July 2018 Revised 12 November 2018 Accepted 15 November 2018 Published 8 January 2019
Academic Editor Maria E Fantacci
Copyright copy 2019 K Senthil Kumar et al +is is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work isproperly cited
+e objective of this paper is to explore an expedient image segmentation algorithm for medical images to curtail the physiciansrsquointerpretation of computer tomography (CT) scan images Modern medical imaging modalities generate large images that areextremely grim to analyze manually +e consequences of segmentation algorithms rely on the exactitude and convergence timeAt this moment there is a compelling necessity to explore and implement new evolutionary algorithms to solve the problemsassociated with medical image segmentation Lung cancer is the frequently diagnosed cancer across the world among men Earlydetection of lung cancer navigates towards apposite treatment to save human lives CT is one of the modest medical imagingmethods to diagnose the lung cancer In the present study the performance of five optimization algorithms namely k-meansclustering k-median clustering particle swarm optimization inertia-weighted particle swarm optimization and guaranteedconvergence particle swarm optimization (GCPSO) to extract the tumor from the lung image has been implemented andanalyzed +e performance of median adaptive median and average filters in the preprocessing stage was compared and it wasproved that the adaptive median filter is most suitable for medical CT images Furthermore the image contrast is enhanced byusing adaptive histogram equalization +e preprocessed image with improved quality is subject to four algorithms +e practicalresults are verified for 20 sample images of the lung using MATLAB and it was observed that the GCPSO has the highest accuracyof 9589
1 Introduction
Lung cancer also known as lung carcinoma is a malignanttumor characterized by uncontrolled growth of the cell intissues of the lung It is mandatory to treat this to avoidspreading its growth by metastasis to other parts of thebody Most cancers that start in the lung are carcinomas+e two main types are small-cell lung carcinoma and non-small-cell lung carcinoma [1] Long-period tobaccosmoking is the primary factor for 85 of lung cancers [2]About 10ndash15 of cases occur in people who have neversmoked but due to air pollution secondhand smokingasbestos and radon gas Computer tomography (CT) andradiographs are the conventional methods to detect thepresence of lung cancer +e diagnosis is confirmed by
biopsy which is usually performed by bronchoscopy or CTscan +e cause of cancer-related death among men ismainly due to lung cancer Hence it is essential to de-termine a new robust method to diagnose the lung cancer atan earlier stage [3] For the present study 20 lung imagesamples and four algorithms have been taken for analysis Itwas proved that the combination of adaptive median filteradaptive histogram equalization and guaranteed conver-gence particle swarm optimization- (GCPSO-) based al-gorithm has more accurate results among others
2 Methods
In medical image segmentation the accuracy is foremostimportant as it deals with human lives It is highly crucial to
HindawiComputational and Mathematical Methods in MedicineVolume 2019 Article ID 4909846 16 pageshttpsdoiorg10115520194909846
Preprocessing
Input CTimage
Median filter
Average filter
Adaptivemedian filter
Adaptivehistogram
equalization
Segmentationand tumorextraction
Guaranteedconvergence PSO
Inertia-weightedPSO
Particle swarmoptimization
k-meansclustering
Outputimage
Segmentation
Figure 1 Process flow diagram of the projected method
(1) Assume the input matrix ldquoArdquo which has M rows and N columns(2) Construct a matrix with M + 2 rows and N + 2 columns by appending zeros to sides of the input matrix(3) Take a mask of size 3 times 3(4) Place the mask on the first element ie element on the first row and first column of matrix ldquoArdquo(5) Select all the elements listed by the mask and sort them in ascending order(6) Take the median value (center element) from the sorted array and replace the element A(1 1) by the median value(7) Slide the mask to the next element(8) Repeat the steps from 4 to 7 until all the elements of matrix ldquoArdquo are replaced by their corresponding median value
ALGORITHM 1 Median filter
(1) Assume the input matrix ldquoArdquo which has M rows and N columns(2) Construct a matrix with M + 2 rows and N + 2 columns by appending zeros to sides of the input matrix(3) Take a mask of size 3 times 3(4) Place the mask on the first element ie element on the first row and first column of matrix ldquoArdquo(5) Select all the elements listed by the mask and find the average(6) Take the mean value from the sorted array and replace the element A(1 1) by the median value(7) Slide the mask to the next element(8) Repeat the steps from 4 to 7 until all the elements of matrix ldquoArdquo are replaced by their corresponding median value
ALGORITHM 2 Median filter
(1) Obtain the histogram for the input image and find the probability mass function(2) Find the cumulative distributive function from that find the CDF according to gray levels(3) Find the new gray levels by using the following equation
CDFNew CDF lowast (number of gray levels minus 1)(4) Map the new gray levels into a total number of pixels and plot the modified histogram
ALGORITHM 3 Histogram equalization
(1) Select the cluster centers Let them be ldquoCrdquo(2) Calculate the Euclidean distance(3) Take each and every pixel and assign them into the appropriate cluster if the Euclidean distance is minimum between the cluster
and pixel(4) Once the segregation is completed for all the pixels recalculate the new cluster center using the following formula
vi (1ci)1113936ci
j1xi
(5) Repeat the steps from 2 to 4 for some number of iterations or until a certain condition is encountered
ALGORITHM 4 k-Means clustering
2 Computational and Mathematical Methods in Medicine
(1) Select the random cluster centers Let the number of cluster centers be ldquoCrdquo(2) Calculate the Euclidean distance(3) Take each and every pixel and assign them into the appropriate cluster if the Euclidean distance is minimum between the cluster
and pixel(4) Once the segregation is completed for all the pixels recalculate the new cluster center using the median value instead of using a
squared formula(5) Repeat the steps from 2 to 4 for some number of iterations or until a certain condition is encountered
ALGORITHM 5 k-Median clustering
(1) Initialize the velocity and position of all the particles with random values(2) Define a fitness function(3) Find the fitness value for each particle(4) Compare the fitness value with the best fitness If the fitness values are better then set the current value as new pbest(5) Repeat steps from 3 to 5 for each particle(6) Update the velocity using equation (1)(7) Upgrade the position(8) Update gbest(9) Repeat steps from 7 to 9 until certain conditions are encountered or for the predefined number of iterations
ALGORITHM 6 Particle swarm optimization [11 13]
Initialization
(1) Initialize the number of clusters and number of iterations(2) Initialize sc fc numSuccess 0 and numFailures 0(3) Define a fitness function
Clustering
(4) Find the fitness value for each particle(5) Update the local best solution obtained so far(6) Repeat steps 4 and 5 for the predefined number of iterations(7) Update velocity and position of each particle for the current global best particle
Selection step
(8) Execute the selection operator(9) If any local best position yi has changed perform the clustering algorithm Otherwise end the algorithm
ALGORITHM 7 GCPSO algorithm [15]
Speckle suppression index (SSI) Speckle suppression and mean preservation index (SMPI)
Q = 1+ Mean(Io) ndash Mean(If)
SSI = lowast
lowastMean(If)
Mean(Io)
Var(Io)
Var(If)
SMPI = QVar(If)
Var(Io)
Figure 2 Performance measures of the filter
Computational and Mathematical Methods in Medicine 3
True positive (tp)Pixels correctly segmented
as foreground
True negative (tn)Pixels correctly detected as
background
False positive (fp)Pixels falsely segmented as
foreground
False negative (fn)Pixels falsely detected as
background
AccuracyA degree of measure to
state the correctness of aprocess
tp + tnAccuracy =
tp + tn + fp + fn
tpr =tp
tp + fn
fnr =fn
fn + tpfpr =
fpfp + tn
tnr =tn
tn + fp
Figure 3 Performance measures for the medical image segmentation
Table 1 SSI and SMPI values of input images
Sample imagesSSI SMPI
Mean filter Median filter Adaptive median filter Mean filter Median filter Adaptive median filterImage 1 09621 08208 08086 09857 09788 09638Image 2 09658 08232 08087 09895 0959 09452Image 3 09588 08209 08091 09883 09799 09696Image 4 09671 08080 07937 09958 09836 09703Image 5 09705 08220 08078 09851 09833 09706Image 6 09708 08218 07900 09948 09775 09457Image 7 09660 08202 08067 09979 09608 09464Image 8 09640 08265 08154 09922 09622 09493Image 9 09638 08272 08141 09990 09716 09576Image 10 09644 08238 08112 09944 09804 09659Image 11 09639 08231 08122 09765 09788 09643Image 12 09642 08289 08152 10012 09826 09721Image 13 09648 08239 08135 09920 09782 09674Image 14 09564 08242 08098 09888 09767 09648Image 15 09573 08208 08084 10005 09785 09636Image 16 09631 08242 08095 09912 09755 09613Image 17 09919 08239 08352 09722 09770 09882Image 18 09912 07983 07857 10003 09808 09696Image 19 09921 08020 07884 10037 09838 09706Image 20 09939 08085 07690 09968 09741 09432
075
08
085
09
095
1
105
SSI v
alue
Sample images
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Mean filterMedian filterAdaptive median filter
Figure 4 Comparative results of SSI values
4 Computational and Mathematical Methods in Medicine
Sample images
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Mean filterMedian filterAdaptive median filter
093094095096097098099
1101
SMPI
val
ue
Figure 5 Comparative results of SMPI values
Table 2 Statistical results from the k-means algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 878783 891554 108446 121217 885937Image 2 866527 895874 104126 133473 882682Image 3 838975 873900 126100 161025 857501Image 4 826502 857011 142989 173498 842186Image 5 835680 842582 157418 164320 839216Image 6 827250 824643 175654 172750 825795Image 7 811893 790554 209446 188107 801519Image 8 802543 777549 222451 197457 790656Image 9 817874 784139 215861 182126 801606Image 10 804304 774794 225206 195696 790378Image 11 817725 780352 219648 182275 799755Image 12 840795 788912 211088 159205 815238Image 13 816145 787989 212011 183855 802806Image 14 798951 786152 213848 201049 793023Image 15 809012 784626 215374 190988 797600Image 16 801249 781480 218520 188751 792121Image 17 801220 781687 218318 198780 792229Image 18 782509 835148 164852 217491 807020Image 19 787041 837431 162569 212959 810816Image 20 767118 843245 156755 232882 801831
Table 3 Statistical results from the k-median clustering segmentation algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 879631 906864 93136 123069 893672Image 2 865908 901719 98281 134092 885622Image 3 833821 889051 110949 166179 862969Image 4 820844 863637 136363 179156 842695Image 5 832410 857769 142231 167590 845294Image 6 825053 842412 157588 174947 833654Image 7 811107 804281 195719 188893 807832Image 8 801857 794033 205967 198143 798186Image 9 821213 797647 202353 178787 809977Image 10 806627 791577 208423 193373 799611Image 11 820209 791588 208412 179791 806621Image 12 843809 804121 195879 156191 824514Image 13 820487 803496 196504 179513 812545Image 14 804506 794375 205625 195494 799876
Computational and Mathematical Methods in Medicine 5
Table 4 Statistical results from the PSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 875413 901196 98804 124587 890624Image 2 859612 859612 85479 85479 889689Image 3 827919 898314 101686 172081 864850Image 4 811271 887838 112162 188729 849967Image 5 826343 873995 126005 173657 850299Image 6 818996 852900 147100 181004 835581Image 7 817281 800949 199051 182719 809438Image 8 804182 800721 199279 195818 802571Image 9 822573 811450 188550 177427 817340Image 10 808521 803433 196567 191479 806182Image 11 822198 809837 190163 177802 816421Image 12 846322 815070 184930 153678 831347Image 13 826283 811153 188847 173617 819351Image 14 809338 809090 190910 190662 809226Image 15 818790 811729 188271 181210 815586Image 16 808120 814222 185778 191880 810836Image 17 808582 818136 181864 191418 812824Image 18 791387 850084 149916 208613 818114Image 19 794655 851570 148430 205345 820954Image 20 777826 853446 146554 222174 811744
Table 5 Statistical results from the IWPSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 874649 903272 96728 125351 889813Image 2 861950 898126 101874 138050 881810Image 3 829347 889622 110378 170653 861018Image 4 814285 872013 127987 185715 843584Image 5 829023 863940 136060 170977 846631Image 6 821065 845145 154855 178935 832876Image 7 821361 791744 208256 178639 807064Image 8 803274 803855 196145 196726 803544Image 9 824185 803924 196076 175815 814631Image 10 810769 795965 204035 189231 803943Image 11 822198 797401 202599 174299 812411Image 12 846322 811390 188610 152172 830334Image 13 826283 807231 173596 173596 818905Image 14 809338 808231 191769 190328 809025Image 15 818790 815899 184101 182694 816669Image 16 808120 814173 185827 191734 810896Image 17 808582 810677 189323 190166 810209Image 18 791387 849622 150378 208310 818081Image 19 794655 848219 151781 203936 820213Image 20 777684 854281 145719 222316 812033
Table 3 Continued
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 15 813002 800536 199464 186998 807248Image 16 801942 800503 199497 198058 801291Image 17 802984 803756 196244 197016 803332Image 18 786327 855226 144774 213673 817792Image 19 789322 852163 147837 210678 818439Image 20 772752 858000 142000 227248 810973
6 Computational and Mathematical Methods in Medicine
Table 6 Statistical results from the GCPSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 916158 999999 00001 83842 958079Image 2 909563 999999 00001 90437 954782Image 3 888404 999999 00001 111592 944204Image 4 872946 999999 00001 127054 936473Image 5 873583 999999 00001 126417 936792Image 6 861567 999999 00001 138433 930784Image 7 834867 999999 00001 165133 917434Image 8 831082 999999 00001 168918 915541Image 9 842907 999999 00001 157093 921453Image 10 831917 999999 00001 168083 915958Image 11 842122 999999 00001 157878 921061Image 12 857977 999999 00001 142023 928988Image 13 846397 999999 00001 153603 923198Image 14 837442 999999 00001 162558 918721Image 15 843299 999999 00001 156701 921649Image 16 838867 999999 00001 161133 919433Image 17 839061 999999 00001 160939 919531Image 18 846836 999999 00001 153164 923418Image 19 849324 999999 00001 150676 924662Image 20 839867 999999 00001 160124 919938
6500
7000
7500
8000
8500
9000
9500
True
pos
itive
rate
k-meansk-mediansPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
IWPSOGCPSO
Figure 6 Comparative results of the true positive rate value
75
80
85
90
95
100
True
neg
ativ
e rat
e
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 7 Comparative results of the true negative rate value
Computational and Mathematical Methods in Medicine 7
72
77
82
87
92
97
Acc
urac
y
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 10 Comparative results of accuracy
0
5
10
15
20
25
False
pos
itive
rate
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 8 Comparative results of the false positive rate value
0
5
10
15
20
25
False
neg
ativ
e rat
e
k-meansk-mediansPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
IWPSOGCPSO
Figure 9 Comparative results of the false negative rate value
8 Computational and Mathematical Methods in Medicine
Table 7 Statistical comparative result of accuracy
Images k-Means k-Median PSO IWPSO GCPSOImage 1 885937 893672 890624 889813 958079Image 2 882682 885622 889689 881810 954782Image 3 857501 862969 864850 861018 944204Image 4 842186 842695 849967 843584 936473Image 5 839216 845294 850299 846631 936792Image 6 825795 833654 835581 832876 930784Image 7 801519 807832 809438 807064 917434Image 8 790656 798186 802571 803544 915541Image 9 801606 809977 817340 814631 921453Image 10 790378 799611 806182 803943 915958Image 11 799755 806621 816421 812411 921061Image 12 815238 824514 831347 830334 928988Image 13 802806 812545 819351 818905 923198Image 14 793023 799876 809226 809025 918721Image 15 797600 807248 815586 816669 921649Image 16 792121 801291 810836 810896 919433Image 17 792229 803332 812824 810209 919531Image 18 807020 817792 818114 818081 923418Image 19 810816 818439 820954 820213 924662Image 20 801831 810973 811744 812033 919938
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 1
Image 2
Image 3
Image 4
Image 5
Image 6
Image 7
Image 8
Image 9
Image 10
(a)
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
(b)
Figure 11 Resultant images after preprocessing
Computational and Mathematical Methods in Medicine 9
eradicate the incidence of noise content and to improve theimage quality before an examination [4]+is part of work isknown as preprocessing In the preprocessing stage noiseremoval and contrast enhancement are two primary steps Inthe present study the performance results of medianadaptive median and average filters to isolate the presence ofspeckle noise have been compared +e coding for the samehas been implemented using MATLAB Furthermore theimage quality and visual appearance are improved byadaptive histogram equalization+e second stage of work issegmentation +is stage consists of applying five methodsnamely k-means k-median particle swarm optimization(PSO) inertia-weighted particle swarm optimization(IWPSO) and GCPSO +e tumor portion was extractedfrom the segmented results of the above-said five methodsand compared with manual extraction+e results show thatthe GCPSO-based segmentation has more accuracy than theothers Figure 1 depicts the process of operation for thepresent study
21 Median and Adaptive Median Filters +e median filterremoves the noise and retains the sharpness of the imageAccordance to the name each pixel is replaced by themedian value from the neighborhood pixels A 3 times 3 windowis used in this filter [5] +is is one of the best filters amongconventional filters which remove the speckle noise +esteps followed to construct the median filter are given inAlgorithm 1
Spatial processing to preserve the edge detail and toeliminate nonimpulsive noise by the adaptive median filterplays a vital role +e small structure in the image and edgesare retained by the adaptive median filter In the adaptivemedian filter the window size varies with respect to eachpixel
22 Average Filter +is is a simple filter which removesthe spatial noise from a digital image +e presence ofspatial noise is mainly due to the data acquisition process
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 12 Resultant images by k-means clustering
10 Computational and Mathematical Methods in Medicine
+e neighborhood mean value is measured for each andevery pixel and is replaced by the corresponding meanvalue +is process is repeated for every pixel in the image[5] All the pixels in the digital image are modified bysliding the operator over the entire range of pixels +esteps followed for the average filter are given inAlgorithm 2
23 Histogram Equalization Image enhancement is thetechnique which is used to improve the image quality Forbetter understanding and analysis it is mandatory to en-hance the contrast of medical images +e conventionalmethod used for this operation is histogram equalization Aminor adjustment on the intensity of image pixels is donein this method Each pixel is mapped to intensity pro-portional to its rank in the surrounding pixels +e stepsfollowed for histogram equalization are given in Algo-rithm 3 [6]
24 k-Means Clustering Algorithm +e simplest and con-ventional method in cluster analysis is the k-means clus-tering algorithm+is algorithm segregates the given datasetinto two or more clusters [7] +e accuracy of this methodcompletely depends on the selection of the cluster center Itis mandatory to select the optimum cluster center to get abetter result +e Euclidean distance is the general measureto segregate the dataset [8] Pixels are assigned to an indi-vidual cluster based on the Euclidean distance +e objectivefunction used in this algorithm is
J(v) 1113944C
i11113944
Ci
j1xi minus vj
1113874 11138752 (1)
where xi are the pixels vj are the cluster centers xi minus vj isthe Euclidean distance between xi and vjCi is the number ofdata points for the ith cluster and C is the number of clustercenters [9] +e steps followed for k-means clustering aregiven in Algorithm 4
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 13 Resultant images by k-median clustering
Computational and Mathematical Methods in Medicine 11
25 k-Median Clustering Algorithm +is is also a clusteringalgorithm slightly modified from the k-means algorithm Incentroid calculation instead of calculating the mean valuethe median value is considered +is algorithm significantlyreduces the error since there is no squared operation as inthe calculation of the Euclidean distance +e clustersformed by this method are more compact As an alternatethis approach uses the Lloyd-style iteration +e steps fol-lowed for k-median clustering are given in Algorithm 5 [10]
26 Particle Swarm Optimization PSO is a metaheuristicalgorithm used efficiently in medical image analysis [11] Itmimics the social behavior of the birds searching for food [12]+e fundamental idea of PSO is sharing and communicatingthe information In this approach each particle has initialposition and velocity Based on the fitness value the velocity
and position are updated +e relevant two equations in PSOto update the position and velocity are as follows [11 12]
v(t + 1) v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minus x(t)]
x(t + 1) x(t) + v(t + 1)
(2)
where r1 and r2 are the random numbers and the accel-eration coefficients c1 and c2 are two positive constants+e success of PSO relies on the fitness function +efollowing fitness function has been used for the presentstudy
maximizef 1113944n
i1
intercluster distanceintracluster distance
(3)
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 14 Resultant images by the PSO algorithm
12 Computational and Mathematical Methods in Medicine
where n is the number of clusters +e steps followed for theparticle swarm optimization are shown in Algorithm 6
27 Inertia-Weighted Particle Swarm Optimization +eexploration and exploitation in PSO are based on the inertiaweight +e basic PSO presented by Eberhart and Kennedyin 1995 has no inertia weight In 1998 Shi and Eberhartintroduced the concept of inertia weight by adding constantinertia weight +ey stated that a significant inertia weightfacilitates a global search while a small inertia weight fa-cilitates a local search [14] +is enhances the convergencerate and reduces the number of iterations Inertia weight lessthan 1 in general improves the results +e used methodimproves the convergence rate and saves the time taken andsome iterations
+e resulting velocity update equation becomes
v(t + 1) wlowast v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minusx(t)](4)
where w is the inertia weight with constant inertia weight w
07 and random inertia weight w 05 + rand()2
28 Guaranteed Convergence Particle Swarm Optimization+e GCPSO focuses on a new particle which deals with thecurrent best position in the region In this task this particleis treated as a member of the swarm and the velocity updateequation for this new particle is given as follows [15]
vφ(t + 1) xφ(t) + pbest(t) + ωvφ(t) + ρ(t)(1minus 2r) (5)
+e search ability is increased by the social part +is willimprove the random search in the area around the gbestposition +e random vector and diameter of the search area
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 15 Resultant images by IWPSO algorithm clustering
Computational and Mathematical Methods in Medicine 13
are r and ρ(t) respectively +e range of the random vectorlies between 0 and 1 +e diameter of the search area can beupdated using the following equation
ρ(t + 1)
2ρ(t) successesgt sc
115
1113874 1113875ρ(t) failuresgt fc
ρ(t) otherwise
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎩
(6)
where the terms successes and failures are defined asthe number of consecutive successes and failures re-spectively +e threshold parameters sc and fc are de-termined empirically Since it is hard to obtain a bettervalue in only a few iterations in a high-dimensional searchspace the recommended values are thus sc 15 and fc 5
On some benchmark tests the GCPSO has shown anexcellent performance of locating the minimal of a spaceafter unimodal with only a small amount of particles +esteps to be followed for the GCPSO are shown inAlgorithm 7
3 Performance Measures
Certain performance measures are used to evaluate theresults obtained from medical image segmentation +e listof performance measures used to assess the filter operation isshown in Figure 2 [16] Let If be the image after noise re-duction and I0 be the noisy image
Performance measures used for the evaluation of theresults of the segmentation algorithm are given in Figure 3[17]
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 16 Resultant images by GCPSO algorithm clustering
14 Computational and Mathematical Methods in Medicine
4 Results and Discussion
+e used methods are practically implemented usingMATLAB coding and the results were verified
In the preprocessing stage a comparison was donebetween the performance of median adaptive median andmean filters +e SSI and SMPI values are shown in Table 1and Figures 4 and 5 From the results it is evident that theadaptive median filter has accurate characteristics than themean and median filters for medical image segmentation
+e segmentation accuracy was measured using the truepositive rate true negative rate false positive rate and falsenegative rate by comparing the results from the algorithmwithmanual segmentation results +e practical results of the k-means clustering segmentation algorithm are shown in Table 2
+e practical results of the k-median clustering seg-mentation algorithm are shown in Table 3
+e practical results of the PSO-based segmentationalgorithm are shown in Table 4
+e practical results of the IWPSO segmentation algo-rithm are shown in Table 5
+e practical results of the GCPSO segmentation algo-rithm are shown in Table 6
+e graphical view of the comparison of the true positiverate true negative rate false positive rate and false negativerate for the algorithms used is shown in Figures 6ndash9 It isproved that the true positive and true negative rates are highand false positive and false negative rates are low for theGCPSO algorithm
+e comparative evaluation based on the accuracy of thesegmentation is shown in Table 7 and Figure 10 +e resultsindicate that the GCPSO-based technique has the highestaverage value of accuracy than the other methods
+e resultant images after preprocessing are shown inFigures 11(a) and 11(b)
+e resultant images after segmentation using k-meansclustering are shown in Figure 12
+e resultant images after segmentation using k-medianclustering are shown in Figure 13
+e resultant images after segmentation using the PSOalgorithm are shown in Figure 14
+e resultant images after segmentation using theIWPSO algorithm are shown in Figure 15
+e resultant images after segmentation using theGCPSO algorithm are shown in Figure 16
In an earlier research lung cancer detection was doneusing PSO genetic optimization and SVM algorithm withthe Gabor filter and produced an accuracy of 895 [18]+emethod to detect lung cancer by means of K-NN classifi-cation using the genetic algorithm produced a maximumaccuracy of 90 [19]+e comparative results with respect tothe above-said methods are shown in Table 8
+e graphical comparative analysis between the used andexisting methods is shown in Figure 17
5 Conclusion
In this study various optimization algorithms have beenevaluated to detect the tumor Medical images often need
preprocessing before being subjected to statistical analysis+e adaptive median filter has better results than medianand mean filters because the speckle suppression index andspeckle and mean preservation index values are lower for theadaptive median filter Comparing the five algorithms theaccuracy of the tumor extraction is improved in GCPSOwith the highest accuracy of 958079 and it obtained above90 of precision in all the 20 images It is more accuratewhen compared to the previous method which had an ac-curacy of 90 in 4 out of 10 datasets only In future studiesthe use of more number of optimization algorithms will beincluded to improve the accuracy
Data Availability
+eCTimages data used to support the findings of this studyhave been deposited in the LungCT-Diagnosis repository(doiorg107937K9TCIA2015A6V7JIWX)
Conflicts of Interest
+e authors declare that there are no conflicts of interestregarding the publication of this paper
References
[1] A A Brindha S Indirani and A Srinivasan ldquoLung cancerdetection using SVM algorithm and optimization techniquesrdquoJournal of Chemical and Pharmaceutical Sciences vol 9 no 42016
[2] M Kurkure and A +akare ldquoIntroducing automated systemfor lung cancer detection using Evolutionary ApproachrdquoInternational Journal of Engineering and Computer Sciencevol 5 no 5 pp 16736ndash16739 2016
[3] B Rani A K Goel and R Kaur ldquoA modified approach forlung cancer detection using bacterial forging optimization
868890929496
PSO GA and SVM algorithm
K-NNclassification
using GA
Proposed GCPSOmethod
8950 90
9581
Acc
urac
y
Various algorithms
Comparison between existing and projected methods
Figure 17 Graphical view of accuracy
Table 8 Comparative analysis of accuracy of the projected methodwith various methods
Various methods Accuracy ()PSO GA and SVM algorithm [18] 8950K-NN classification using GA [19] 9000Projected GCPSO method 9581
Computational and Mathematical Methods in Medicine 15
algorithmrdquo International Journal of Scientific Research En-gineering and Technology vol 5 no 1 2016
[4] N Panpaliya N Tadas S Bobade R Aglawe andA Gudadhe ldquoA survey on early detection and prediction oflung cancerrdquo International Journal of Computer Science andMobile Computing vol 4 no 1 pp 175ndash184 2015
[5] G Gupta ldquoAlgorithm for image processing using improvedmedian filter and comparison of mean median and improvedmedian filterrdquo International Journal of Soft Computing andEngineering vol 1 no 5 2011
[6] Tutorial point with digital image processing httpwwwtutorialspointcomdiphistogram_equalizationhtm
[7] K Venkatalakshmi and S M Shalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and K-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Melbourne VIC Australia December2005
[8] K Venkatalakshmi and S M Shalinie ldquoMultispectral imageclassification using modified k-means clusteringrdquo In-ternational Journal on Neural and Mass-Parallel Computingand Information Systems vol 17 no 2 pp 113ndash120 2007
[9] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercy Shalinie ldquoMultispectral image clustering using en-hanced genetic k-means algorithmrdquo Information TechnologyJournal vol 6 no 4 pp 554ndash560 2007
[10] P I Dalatu ldquoTime complexity of k-means and k-mediansclustering algorithms in outliers detectionrdquo Global Journal ofPure and Applied Mathematics vol 12 no 5 pp 4405ndash44182016
[11] K Senthilkumar K Venkatalakshmi K Karthikeyan andP Kathirkamasundari ldquoAn efficient method for segmentingdigital image using a hybrid model of particle swarm opti-mization and artificial bee colony algorithmrdquo InternationalJournal of Applied Engineering Research vol 10 pp 444ndash4492015
[12] K Venkatalakshmi and S Mercyshalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and k-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Bangalore India December 2005
[13] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercyshalinie ldquoA customized particle swarm optimizationfor classification of multispectral imagery based on featurefusionrdquo International Arab Journal of Information Technologyvol 5 no 4 pp 327ndash333 2008
[14] J C Bansal and P K Singh ldquoInertia weight strategies inparticle swarm optimizationrdquo in Proceedings of IEEE WorldCongress on Nature and Biologically Inspired Computing pp640ndash647 Salamanca Spain October 2011
[15] P K Patel V Sharma and K Gupta ldquoGuaranteed conver-gence particle swarm optimization using Personal Bestrdquo In-ternational Journal of Computer Applications vol 73 no 7pp 6ndash10 2013
[16] X Wang L Ge and X Li ldquoEvaluation of filters for Envi-satAsar speckle suppression in pasture areardquo ISPRS Annals ofthe Photogrammetry Remote Sensing and Spatial InformationSciences vol 1ndash7 pp 341ndash346 2012
[17] H Nagaveena D Devaraj and S C Prasanna KumarldquoVessels segmentation in diabetic retinopathy by adaptivemedian thresholdingrdquo International Journal of Science andTechnology vol 1 no 1 pp 17ndash22 2013
[18] A Asuntha N Singh and A Srinivasan ldquoPSO genetic op-timization and SVM algorithm used for lung cancer
detectionrdquo Journal of Chemical and Pharmaceutical Researchvol 8 no 6 pp 351ndash359 2016
[19] P Bhuvaneswari and A Brintha+erese ldquoDetection of cancerin lung with K-NN classification using genetic algorithmrdquoInternational Conference on Nanomaterials and Technologiesvol 10 pp 433ndash440 2014
16 Computational and Mathematical Methods in Medicine
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom
Preprocessing
Input CTimage
Median filter
Average filter
Adaptivemedian filter
Adaptivehistogram
equalization
Segmentationand tumorextraction
Guaranteedconvergence PSO
Inertia-weightedPSO
Particle swarmoptimization
k-meansclustering
Outputimage
Segmentation
Figure 1 Process flow diagram of the projected method
(1) Assume the input matrix ldquoArdquo which has M rows and N columns(2) Construct a matrix with M + 2 rows and N + 2 columns by appending zeros to sides of the input matrix(3) Take a mask of size 3 times 3(4) Place the mask on the first element ie element on the first row and first column of matrix ldquoArdquo(5) Select all the elements listed by the mask and sort them in ascending order(6) Take the median value (center element) from the sorted array and replace the element A(1 1) by the median value(7) Slide the mask to the next element(8) Repeat the steps from 4 to 7 until all the elements of matrix ldquoArdquo are replaced by their corresponding median value
ALGORITHM 1 Median filter
(1) Assume the input matrix ldquoArdquo which has M rows and N columns(2) Construct a matrix with M + 2 rows and N + 2 columns by appending zeros to sides of the input matrix(3) Take a mask of size 3 times 3(4) Place the mask on the first element ie element on the first row and first column of matrix ldquoArdquo(5) Select all the elements listed by the mask and find the average(6) Take the mean value from the sorted array and replace the element A(1 1) by the median value(7) Slide the mask to the next element(8) Repeat the steps from 4 to 7 until all the elements of matrix ldquoArdquo are replaced by their corresponding median value
ALGORITHM 2 Median filter
(1) Obtain the histogram for the input image and find the probability mass function(2) Find the cumulative distributive function from that find the CDF according to gray levels(3) Find the new gray levels by using the following equation
CDFNew CDF lowast (number of gray levels minus 1)(4) Map the new gray levels into a total number of pixels and plot the modified histogram
ALGORITHM 3 Histogram equalization
(1) Select the cluster centers Let them be ldquoCrdquo(2) Calculate the Euclidean distance(3) Take each and every pixel and assign them into the appropriate cluster if the Euclidean distance is minimum between the cluster
and pixel(4) Once the segregation is completed for all the pixels recalculate the new cluster center using the following formula
vi (1ci)1113936ci
j1xi
(5) Repeat the steps from 2 to 4 for some number of iterations or until a certain condition is encountered
ALGORITHM 4 k-Means clustering
2 Computational and Mathematical Methods in Medicine
(1) Select the random cluster centers Let the number of cluster centers be ldquoCrdquo(2) Calculate the Euclidean distance(3) Take each and every pixel and assign them into the appropriate cluster if the Euclidean distance is minimum between the cluster
and pixel(4) Once the segregation is completed for all the pixels recalculate the new cluster center using the median value instead of using a
squared formula(5) Repeat the steps from 2 to 4 for some number of iterations or until a certain condition is encountered
ALGORITHM 5 k-Median clustering
(1) Initialize the velocity and position of all the particles with random values(2) Define a fitness function(3) Find the fitness value for each particle(4) Compare the fitness value with the best fitness If the fitness values are better then set the current value as new pbest(5) Repeat steps from 3 to 5 for each particle(6) Update the velocity using equation (1)(7) Upgrade the position(8) Update gbest(9) Repeat steps from 7 to 9 until certain conditions are encountered or for the predefined number of iterations
ALGORITHM 6 Particle swarm optimization [11 13]
Initialization
(1) Initialize the number of clusters and number of iterations(2) Initialize sc fc numSuccess 0 and numFailures 0(3) Define a fitness function
Clustering
(4) Find the fitness value for each particle(5) Update the local best solution obtained so far(6) Repeat steps 4 and 5 for the predefined number of iterations(7) Update velocity and position of each particle for the current global best particle
Selection step
(8) Execute the selection operator(9) If any local best position yi has changed perform the clustering algorithm Otherwise end the algorithm
ALGORITHM 7 GCPSO algorithm [15]
Speckle suppression index (SSI) Speckle suppression and mean preservation index (SMPI)
Q = 1+ Mean(Io) ndash Mean(If)
SSI = lowast
lowastMean(If)
Mean(Io)
Var(Io)
Var(If)
SMPI = QVar(If)
Var(Io)
Figure 2 Performance measures of the filter
Computational and Mathematical Methods in Medicine 3
True positive (tp)Pixels correctly segmented
as foreground
True negative (tn)Pixels correctly detected as
background
False positive (fp)Pixels falsely segmented as
foreground
False negative (fn)Pixels falsely detected as
background
AccuracyA degree of measure to
state the correctness of aprocess
tp + tnAccuracy =
tp + tn + fp + fn
tpr =tp
tp + fn
fnr =fn
fn + tpfpr =
fpfp + tn
tnr =tn
tn + fp
Figure 3 Performance measures for the medical image segmentation
Table 1 SSI and SMPI values of input images
Sample imagesSSI SMPI
Mean filter Median filter Adaptive median filter Mean filter Median filter Adaptive median filterImage 1 09621 08208 08086 09857 09788 09638Image 2 09658 08232 08087 09895 0959 09452Image 3 09588 08209 08091 09883 09799 09696Image 4 09671 08080 07937 09958 09836 09703Image 5 09705 08220 08078 09851 09833 09706Image 6 09708 08218 07900 09948 09775 09457Image 7 09660 08202 08067 09979 09608 09464Image 8 09640 08265 08154 09922 09622 09493Image 9 09638 08272 08141 09990 09716 09576Image 10 09644 08238 08112 09944 09804 09659Image 11 09639 08231 08122 09765 09788 09643Image 12 09642 08289 08152 10012 09826 09721Image 13 09648 08239 08135 09920 09782 09674Image 14 09564 08242 08098 09888 09767 09648Image 15 09573 08208 08084 10005 09785 09636Image 16 09631 08242 08095 09912 09755 09613Image 17 09919 08239 08352 09722 09770 09882Image 18 09912 07983 07857 10003 09808 09696Image 19 09921 08020 07884 10037 09838 09706Image 20 09939 08085 07690 09968 09741 09432
075
08
085
09
095
1
105
SSI v
alue
Sample images
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Mean filterMedian filterAdaptive median filter
Figure 4 Comparative results of SSI values
4 Computational and Mathematical Methods in Medicine
Sample images
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Mean filterMedian filterAdaptive median filter
093094095096097098099
1101
SMPI
val
ue
Figure 5 Comparative results of SMPI values
Table 2 Statistical results from the k-means algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 878783 891554 108446 121217 885937Image 2 866527 895874 104126 133473 882682Image 3 838975 873900 126100 161025 857501Image 4 826502 857011 142989 173498 842186Image 5 835680 842582 157418 164320 839216Image 6 827250 824643 175654 172750 825795Image 7 811893 790554 209446 188107 801519Image 8 802543 777549 222451 197457 790656Image 9 817874 784139 215861 182126 801606Image 10 804304 774794 225206 195696 790378Image 11 817725 780352 219648 182275 799755Image 12 840795 788912 211088 159205 815238Image 13 816145 787989 212011 183855 802806Image 14 798951 786152 213848 201049 793023Image 15 809012 784626 215374 190988 797600Image 16 801249 781480 218520 188751 792121Image 17 801220 781687 218318 198780 792229Image 18 782509 835148 164852 217491 807020Image 19 787041 837431 162569 212959 810816Image 20 767118 843245 156755 232882 801831
Table 3 Statistical results from the k-median clustering segmentation algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 879631 906864 93136 123069 893672Image 2 865908 901719 98281 134092 885622Image 3 833821 889051 110949 166179 862969Image 4 820844 863637 136363 179156 842695Image 5 832410 857769 142231 167590 845294Image 6 825053 842412 157588 174947 833654Image 7 811107 804281 195719 188893 807832Image 8 801857 794033 205967 198143 798186Image 9 821213 797647 202353 178787 809977Image 10 806627 791577 208423 193373 799611Image 11 820209 791588 208412 179791 806621Image 12 843809 804121 195879 156191 824514Image 13 820487 803496 196504 179513 812545Image 14 804506 794375 205625 195494 799876
Computational and Mathematical Methods in Medicine 5
Table 4 Statistical results from the PSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 875413 901196 98804 124587 890624Image 2 859612 859612 85479 85479 889689Image 3 827919 898314 101686 172081 864850Image 4 811271 887838 112162 188729 849967Image 5 826343 873995 126005 173657 850299Image 6 818996 852900 147100 181004 835581Image 7 817281 800949 199051 182719 809438Image 8 804182 800721 199279 195818 802571Image 9 822573 811450 188550 177427 817340Image 10 808521 803433 196567 191479 806182Image 11 822198 809837 190163 177802 816421Image 12 846322 815070 184930 153678 831347Image 13 826283 811153 188847 173617 819351Image 14 809338 809090 190910 190662 809226Image 15 818790 811729 188271 181210 815586Image 16 808120 814222 185778 191880 810836Image 17 808582 818136 181864 191418 812824Image 18 791387 850084 149916 208613 818114Image 19 794655 851570 148430 205345 820954Image 20 777826 853446 146554 222174 811744
Table 5 Statistical results from the IWPSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 874649 903272 96728 125351 889813Image 2 861950 898126 101874 138050 881810Image 3 829347 889622 110378 170653 861018Image 4 814285 872013 127987 185715 843584Image 5 829023 863940 136060 170977 846631Image 6 821065 845145 154855 178935 832876Image 7 821361 791744 208256 178639 807064Image 8 803274 803855 196145 196726 803544Image 9 824185 803924 196076 175815 814631Image 10 810769 795965 204035 189231 803943Image 11 822198 797401 202599 174299 812411Image 12 846322 811390 188610 152172 830334Image 13 826283 807231 173596 173596 818905Image 14 809338 808231 191769 190328 809025Image 15 818790 815899 184101 182694 816669Image 16 808120 814173 185827 191734 810896Image 17 808582 810677 189323 190166 810209Image 18 791387 849622 150378 208310 818081Image 19 794655 848219 151781 203936 820213Image 20 777684 854281 145719 222316 812033
Table 3 Continued
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 15 813002 800536 199464 186998 807248Image 16 801942 800503 199497 198058 801291Image 17 802984 803756 196244 197016 803332Image 18 786327 855226 144774 213673 817792Image 19 789322 852163 147837 210678 818439Image 20 772752 858000 142000 227248 810973
6 Computational and Mathematical Methods in Medicine
Table 6 Statistical results from the GCPSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 916158 999999 00001 83842 958079Image 2 909563 999999 00001 90437 954782Image 3 888404 999999 00001 111592 944204Image 4 872946 999999 00001 127054 936473Image 5 873583 999999 00001 126417 936792Image 6 861567 999999 00001 138433 930784Image 7 834867 999999 00001 165133 917434Image 8 831082 999999 00001 168918 915541Image 9 842907 999999 00001 157093 921453Image 10 831917 999999 00001 168083 915958Image 11 842122 999999 00001 157878 921061Image 12 857977 999999 00001 142023 928988Image 13 846397 999999 00001 153603 923198Image 14 837442 999999 00001 162558 918721Image 15 843299 999999 00001 156701 921649Image 16 838867 999999 00001 161133 919433Image 17 839061 999999 00001 160939 919531Image 18 846836 999999 00001 153164 923418Image 19 849324 999999 00001 150676 924662Image 20 839867 999999 00001 160124 919938
6500
7000
7500
8000
8500
9000
9500
True
pos
itive
rate
k-meansk-mediansPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
IWPSOGCPSO
Figure 6 Comparative results of the true positive rate value
75
80
85
90
95
100
True
neg
ativ
e rat
e
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 7 Comparative results of the true negative rate value
Computational and Mathematical Methods in Medicine 7
72
77
82
87
92
97
Acc
urac
y
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 10 Comparative results of accuracy
0
5
10
15
20
25
False
pos
itive
rate
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 8 Comparative results of the false positive rate value
0
5
10
15
20
25
False
neg
ativ
e rat
e
k-meansk-mediansPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
IWPSOGCPSO
Figure 9 Comparative results of the false negative rate value
8 Computational and Mathematical Methods in Medicine
Table 7 Statistical comparative result of accuracy
Images k-Means k-Median PSO IWPSO GCPSOImage 1 885937 893672 890624 889813 958079Image 2 882682 885622 889689 881810 954782Image 3 857501 862969 864850 861018 944204Image 4 842186 842695 849967 843584 936473Image 5 839216 845294 850299 846631 936792Image 6 825795 833654 835581 832876 930784Image 7 801519 807832 809438 807064 917434Image 8 790656 798186 802571 803544 915541Image 9 801606 809977 817340 814631 921453Image 10 790378 799611 806182 803943 915958Image 11 799755 806621 816421 812411 921061Image 12 815238 824514 831347 830334 928988Image 13 802806 812545 819351 818905 923198Image 14 793023 799876 809226 809025 918721Image 15 797600 807248 815586 816669 921649Image 16 792121 801291 810836 810896 919433Image 17 792229 803332 812824 810209 919531Image 18 807020 817792 818114 818081 923418Image 19 810816 818439 820954 820213 924662Image 20 801831 810973 811744 812033 919938
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 1
Image 2
Image 3
Image 4
Image 5
Image 6
Image 7
Image 8
Image 9
Image 10
(a)
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
(b)
Figure 11 Resultant images after preprocessing
Computational and Mathematical Methods in Medicine 9
eradicate the incidence of noise content and to improve theimage quality before an examination [4]+is part of work isknown as preprocessing In the preprocessing stage noiseremoval and contrast enhancement are two primary steps Inthe present study the performance results of medianadaptive median and average filters to isolate the presence ofspeckle noise have been compared +e coding for the samehas been implemented using MATLAB Furthermore theimage quality and visual appearance are improved byadaptive histogram equalization+e second stage of work issegmentation +is stage consists of applying five methodsnamely k-means k-median particle swarm optimization(PSO) inertia-weighted particle swarm optimization(IWPSO) and GCPSO +e tumor portion was extractedfrom the segmented results of the above-said five methodsand compared with manual extraction+e results show thatthe GCPSO-based segmentation has more accuracy than theothers Figure 1 depicts the process of operation for thepresent study
21 Median and Adaptive Median Filters +e median filterremoves the noise and retains the sharpness of the imageAccordance to the name each pixel is replaced by themedian value from the neighborhood pixels A 3 times 3 windowis used in this filter [5] +is is one of the best filters amongconventional filters which remove the speckle noise +esteps followed to construct the median filter are given inAlgorithm 1
Spatial processing to preserve the edge detail and toeliminate nonimpulsive noise by the adaptive median filterplays a vital role +e small structure in the image and edgesare retained by the adaptive median filter In the adaptivemedian filter the window size varies with respect to eachpixel
22 Average Filter +is is a simple filter which removesthe spatial noise from a digital image +e presence ofspatial noise is mainly due to the data acquisition process
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 12 Resultant images by k-means clustering
10 Computational and Mathematical Methods in Medicine
+e neighborhood mean value is measured for each andevery pixel and is replaced by the corresponding meanvalue +is process is repeated for every pixel in the image[5] All the pixels in the digital image are modified bysliding the operator over the entire range of pixels +esteps followed for the average filter are given inAlgorithm 2
23 Histogram Equalization Image enhancement is thetechnique which is used to improve the image quality Forbetter understanding and analysis it is mandatory to en-hance the contrast of medical images +e conventionalmethod used for this operation is histogram equalization Aminor adjustment on the intensity of image pixels is donein this method Each pixel is mapped to intensity pro-portional to its rank in the surrounding pixels +e stepsfollowed for histogram equalization are given in Algo-rithm 3 [6]
24 k-Means Clustering Algorithm +e simplest and con-ventional method in cluster analysis is the k-means clus-tering algorithm+is algorithm segregates the given datasetinto two or more clusters [7] +e accuracy of this methodcompletely depends on the selection of the cluster center Itis mandatory to select the optimum cluster center to get abetter result +e Euclidean distance is the general measureto segregate the dataset [8] Pixels are assigned to an indi-vidual cluster based on the Euclidean distance +e objectivefunction used in this algorithm is
J(v) 1113944C
i11113944
Ci
j1xi minus vj
1113874 11138752 (1)
where xi are the pixels vj are the cluster centers xi minus vj isthe Euclidean distance between xi and vjCi is the number ofdata points for the ith cluster and C is the number of clustercenters [9] +e steps followed for k-means clustering aregiven in Algorithm 4
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 13 Resultant images by k-median clustering
Computational and Mathematical Methods in Medicine 11
25 k-Median Clustering Algorithm +is is also a clusteringalgorithm slightly modified from the k-means algorithm Incentroid calculation instead of calculating the mean valuethe median value is considered +is algorithm significantlyreduces the error since there is no squared operation as inthe calculation of the Euclidean distance +e clustersformed by this method are more compact As an alternatethis approach uses the Lloyd-style iteration +e steps fol-lowed for k-median clustering are given in Algorithm 5 [10]
26 Particle Swarm Optimization PSO is a metaheuristicalgorithm used efficiently in medical image analysis [11] Itmimics the social behavior of the birds searching for food [12]+e fundamental idea of PSO is sharing and communicatingthe information In this approach each particle has initialposition and velocity Based on the fitness value the velocity
and position are updated +e relevant two equations in PSOto update the position and velocity are as follows [11 12]
v(t + 1) v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minus x(t)]
x(t + 1) x(t) + v(t + 1)
(2)
where r1 and r2 are the random numbers and the accel-eration coefficients c1 and c2 are two positive constants+e success of PSO relies on the fitness function +efollowing fitness function has been used for the presentstudy
maximizef 1113944n
i1
intercluster distanceintracluster distance
(3)
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 14 Resultant images by the PSO algorithm
12 Computational and Mathematical Methods in Medicine
where n is the number of clusters +e steps followed for theparticle swarm optimization are shown in Algorithm 6
27 Inertia-Weighted Particle Swarm Optimization +eexploration and exploitation in PSO are based on the inertiaweight +e basic PSO presented by Eberhart and Kennedyin 1995 has no inertia weight In 1998 Shi and Eberhartintroduced the concept of inertia weight by adding constantinertia weight +ey stated that a significant inertia weightfacilitates a global search while a small inertia weight fa-cilitates a local search [14] +is enhances the convergencerate and reduces the number of iterations Inertia weight lessthan 1 in general improves the results +e used methodimproves the convergence rate and saves the time taken andsome iterations
+e resulting velocity update equation becomes
v(t + 1) wlowast v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minusx(t)](4)
where w is the inertia weight with constant inertia weight w
07 and random inertia weight w 05 + rand()2
28 Guaranteed Convergence Particle Swarm Optimization+e GCPSO focuses on a new particle which deals with thecurrent best position in the region In this task this particleis treated as a member of the swarm and the velocity updateequation for this new particle is given as follows [15]
vφ(t + 1) xφ(t) + pbest(t) + ωvφ(t) + ρ(t)(1minus 2r) (5)
+e search ability is increased by the social part +is willimprove the random search in the area around the gbestposition +e random vector and diameter of the search area
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 15 Resultant images by IWPSO algorithm clustering
Computational and Mathematical Methods in Medicine 13
are r and ρ(t) respectively +e range of the random vectorlies between 0 and 1 +e diameter of the search area can beupdated using the following equation
ρ(t + 1)
2ρ(t) successesgt sc
115
1113874 1113875ρ(t) failuresgt fc
ρ(t) otherwise
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎩
(6)
where the terms successes and failures are defined asthe number of consecutive successes and failures re-spectively +e threshold parameters sc and fc are de-termined empirically Since it is hard to obtain a bettervalue in only a few iterations in a high-dimensional searchspace the recommended values are thus sc 15 and fc 5
On some benchmark tests the GCPSO has shown anexcellent performance of locating the minimal of a spaceafter unimodal with only a small amount of particles +esteps to be followed for the GCPSO are shown inAlgorithm 7
3 Performance Measures
Certain performance measures are used to evaluate theresults obtained from medical image segmentation +e listof performance measures used to assess the filter operation isshown in Figure 2 [16] Let If be the image after noise re-duction and I0 be the noisy image
Performance measures used for the evaluation of theresults of the segmentation algorithm are given in Figure 3[17]
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 16 Resultant images by GCPSO algorithm clustering
14 Computational and Mathematical Methods in Medicine
4 Results and Discussion
+e used methods are practically implemented usingMATLAB coding and the results were verified
In the preprocessing stage a comparison was donebetween the performance of median adaptive median andmean filters +e SSI and SMPI values are shown in Table 1and Figures 4 and 5 From the results it is evident that theadaptive median filter has accurate characteristics than themean and median filters for medical image segmentation
+e segmentation accuracy was measured using the truepositive rate true negative rate false positive rate and falsenegative rate by comparing the results from the algorithmwithmanual segmentation results +e practical results of the k-means clustering segmentation algorithm are shown in Table 2
+e practical results of the k-median clustering seg-mentation algorithm are shown in Table 3
+e practical results of the PSO-based segmentationalgorithm are shown in Table 4
+e practical results of the IWPSO segmentation algo-rithm are shown in Table 5
+e practical results of the GCPSO segmentation algo-rithm are shown in Table 6
+e graphical view of the comparison of the true positiverate true negative rate false positive rate and false negativerate for the algorithms used is shown in Figures 6ndash9 It isproved that the true positive and true negative rates are highand false positive and false negative rates are low for theGCPSO algorithm
+e comparative evaluation based on the accuracy of thesegmentation is shown in Table 7 and Figure 10 +e resultsindicate that the GCPSO-based technique has the highestaverage value of accuracy than the other methods
+e resultant images after preprocessing are shown inFigures 11(a) and 11(b)
+e resultant images after segmentation using k-meansclustering are shown in Figure 12
+e resultant images after segmentation using k-medianclustering are shown in Figure 13
+e resultant images after segmentation using the PSOalgorithm are shown in Figure 14
+e resultant images after segmentation using theIWPSO algorithm are shown in Figure 15
+e resultant images after segmentation using theGCPSO algorithm are shown in Figure 16
In an earlier research lung cancer detection was doneusing PSO genetic optimization and SVM algorithm withthe Gabor filter and produced an accuracy of 895 [18]+emethod to detect lung cancer by means of K-NN classifi-cation using the genetic algorithm produced a maximumaccuracy of 90 [19]+e comparative results with respect tothe above-said methods are shown in Table 8
+e graphical comparative analysis between the used andexisting methods is shown in Figure 17
5 Conclusion
In this study various optimization algorithms have beenevaluated to detect the tumor Medical images often need
preprocessing before being subjected to statistical analysis+e adaptive median filter has better results than medianand mean filters because the speckle suppression index andspeckle and mean preservation index values are lower for theadaptive median filter Comparing the five algorithms theaccuracy of the tumor extraction is improved in GCPSOwith the highest accuracy of 958079 and it obtained above90 of precision in all the 20 images It is more accuratewhen compared to the previous method which had an ac-curacy of 90 in 4 out of 10 datasets only In future studiesthe use of more number of optimization algorithms will beincluded to improve the accuracy
Data Availability
+eCTimages data used to support the findings of this studyhave been deposited in the LungCT-Diagnosis repository(doiorg107937K9TCIA2015A6V7JIWX)
Conflicts of Interest
+e authors declare that there are no conflicts of interestregarding the publication of this paper
References
[1] A A Brindha S Indirani and A Srinivasan ldquoLung cancerdetection using SVM algorithm and optimization techniquesrdquoJournal of Chemical and Pharmaceutical Sciences vol 9 no 42016
[2] M Kurkure and A +akare ldquoIntroducing automated systemfor lung cancer detection using Evolutionary ApproachrdquoInternational Journal of Engineering and Computer Sciencevol 5 no 5 pp 16736ndash16739 2016
[3] B Rani A K Goel and R Kaur ldquoA modified approach forlung cancer detection using bacterial forging optimization
868890929496
PSO GA and SVM algorithm
K-NNclassification
using GA
Proposed GCPSOmethod
8950 90
9581
Acc
urac
y
Various algorithms
Comparison between existing and projected methods
Figure 17 Graphical view of accuracy
Table 8 Comparative analysis of accuracy of the projected methodwith various methods
Various methods Accuracy ()PSO GA and SVM algorithm [18] 8950K-NN classification using GA [19] 9000Projected GCPSO method 9581
Computational and Mathematical Methods in Medicine 15
algorithmrdquo International Journal of Scientific Research En-gineering and Technology vol 5 no 1 2016
[4] N Panpaliya N Tadas S Bobade R Aglawe andA Gudadhe ldquoA survey on early detection and prediction oflung cancerrdquo International Journal of Computer Science andMobile Computing vol 4 no 1 pp 175ndash184 2015
[5] G Gupta ldquoAlgorithm for image processing using improvedmedian filter and comparison of mean median and improvedmedian filterrdquo International Journal of Soft Computing andEngineering vol 1 no 5 2011
[6] Tutorial point with digital image processing httpwwwtutorialspointcomdiphistogram_equalizationhtm
[7] K Venkatalakshmi and S M Shalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and K-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Melbourne VIC Australia December2005
[8] K Venkatalakshmi and S M Shalinie ldquoMultispectral imageclassification using modified k-means clusteringrdquo In-ternational Journal on Neural and Mass-Parallel Computingand Information Systems vol 17 no 2 pp 113ndash120 2007
[9] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercy Shalinie ldquoMultispectral image clustering using en-hanced genetic k-means algorithmrdquo Information TechnologyJournal vol 6 no 4 pp 554ndash560 2007
[10] P I Dalatu ldquoTime complexity of k-means and k-mediansclustering algorithms in outliers detectionrdquo Global Journal ofPure and Applied Mathematics vol 12 no 5 pp 4405ndash44182016
[11] K Senthilkumar K Venkatalakshmi K Karthikeyan andP Kathirkamasundari ldquoAn efficient method for segmentingdigital image using a hybrid model of particle swarm opti-mization and artificial bee colony algorithmrdquo InternationalJournal of Applied Engineering Research vol 10 pp 444ndash4492015
[12] K Venkatalakshmi and S Mercyshalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and k-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Bangalore India December 2005
[13] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercyshalinie ldquoA customized particle swarm optimizationfor classification of multispectral imagery based on featurefusionrdquo International Arab Journal of Information Technologyvol 5 no 4 pp 327ndash333 2008
[14] J C Bansal and P K Singh ldquoInertia weight strategies inparticle swarm optimizationrdquo in Proceedings of IEEE WorldCongress on Nature and Biologically Inspired Computing pp640ndash647 Salamanca Spain October 2011
[15] P K Patel V Sharma and K Gupta ldquoGuaranteed conver-gence particle swarm optimization using Personal Bestrdquo In-ternational Journal of Computer Applications vol 73 no 7pp 6ndash10 2013
[16] X Wang L Ge and X Li ldquoEvaluation of filters for Envi-satAsar speckle suppression in pasture areardquo ISPRS Annals ofthe Photogrammetry Remote Sensing and Spatial InformationSciences vol 1ndash7 pp 341ndash346 2012
[17] H Nagaveena D Devaraj and S C Prasanna KumarldquoVessels segmentation in diabetic retinopathy by adaptivemedian thresholdingrdquo International Journal of Science andTechnology vol 1 no 1 pp 17ndash22 2013
[18] A Asuntha N Singh and A Srinivasan ldquoPSO genetic op-timization and SVM algorithm used for lung cancer
detectionrdquo Journal of Chemical and Pharmaceutical Researchvol 8 no 6 pp 351ndash359 2016
[19] P Bhuvaneswari and A Brintha+erese ldquoDetection of cancerin lung with K-NN classification using genetic algorithmrdquoInternational Conference on Nanomaterials and Technologiesvol 10 pp 433ndash440 2014
16 Computational and Mathematical Methods in Medicine
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom
(1) Select the random cluster centers Let the number of cluster centers be ldquoCrdquo(2) Calculate the Euclidean distance(3) Take each and every pixel and assign them into the appropriate cluster if the Euclidean distance is minimum between the cluster
and pixel(4) Once the segregation is completed for all the pixels recalculate the new cluster center using the median value instead of using a
squared formula(5) Repeat the steps from 2 to 4 for some number of iterations or until a certain condition is encountered
ALGORITHM 5 k-Median clustering
(1) Initialize the velocity and position of all the particles with random values(2) Define a fitness function(3) Find the fitness value for each particle(4) Compare the fitness value with the best fitness If the fitness values are better then set the current value as new pbest(5) Repeat steps from 3 to 5 for each particle(6) Update the velocity using equation (1)(7) Upgrade the position(8) Update gbest(9) Repeat steps from 7 to 9 until certain conditions are encountered or for the predefined number of iterations
ALGORITHM 6 Particle swarm optimization [11 13]
Initialization
(1) Initialize the number of clusters and number of iterations(2) Initialize sc fc numSuccess 0 and numFailures 0(3) Define a fitness function
Clustering
(4) Find the fitness value for each particle(5) Update the local best solution obtained so far(6) Repeat steps 4 and 5 for the predefined number of iterations(7) Update velocity and position of each particle for the current global best particle
Selection step
(8) Execute the selection operator(9) If any local best position yi has changed perform the clustering algorithm Otherwise end the algorithm
ALGORITHM 7 GCPSO algorithm [15]
Speckle suppression index (SSI) Speckle suppression and mean preservation index (SMPI)
Q = 1+ Mean(Io) ndash Mean(If)
SSI = lowast
lowastMean(If)
Mean(Io)
Var(Io)
Var(If)
SMPI = QVar(If)
Var(Io)
Figure 2 Performance measures of the filter
Computational and Mathematical Methods in Medicine 3
True positive (tp)Pixels correctly segmented
as foreground
True negative (tn)Pixels correctly detected as
background
False positive (fp)Pixels falsely segmented as
foreground
False negative (fn)Pixels falsely detected as
background
AccuracyA degree of measure to
state the correctness of aprocess
tp + tnAccuracy =
tp + tn + fp + fn
tpr =tp
tp + fn
fnr =fn
fn + tpfpr =
fpfp + tn
tnr =tn
tn + fp
Figure 3 Performance measures for the medical image segmentation
Table 1 SSI and SMPI values of input images
Sample imagesSSI SMPI
Mean filter Median filter Adaptive median filter Mean filter Median filter Adaptive median filterImage 1 09621 08208 08086 09857 09788 09638Image 2 09658 08232 08087 09895 0959 09452Image 3 09588 08209 08091 09883 09799 09696Image 4 09671 08080 07937 09958 09836 09703Image 5 09705 08220 08078 09851 09833 09706Image 6 09708 08218 07900 09948 09775 09457Image 7 09660 08202 08067 09979 09608 09464Image 8 09640 08265 08154 09922 09622 09493Image 9 09638 08272 08141 09990 09716 09576Image 10 09644 08238 08112 09944 09804 09659Image 11 09639 08231 08122 09765 09788 09643Image 12 09642 08289 08152 10012 09826 09721Image 13 09648 08239 08135 09920 09782 09674Image 14 09564 08242 08098 09888 09767 09648Image 15 09573 08208 08084 10005 09785 09636Image 16 09631 08242 08095 09912 09755 09613Image 17 09919 08239 08352 09722 09770 09882Image 18 09912 07983 07857 10003 09808 09696Image 19 09921 08020 07884 10037 09838 09706Image 20 09939 08085 07690 09968 09741 09432
075
08
085
09
095
1
105
SSI v
alue
Sample images
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Mean filterMedian filterAdaptive median filter
Figure 4 Comparative results of SSI values
4 Computational and Mathematical Methods in Medicine
Sample images
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Mean filterMedian filterAdaptive median filter
093094095096097098099
1101
SMPI
val
ue
Figure 5 Comparative results of SMPI values
Table 2 Statistical results from the k-means algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 878783 891554 108446 121217 885937Image 2 866527 895874 104126 133473 882682Image 3 838975 873900 126100 161025 857501Image 4 826502 857011 142989 173498 842186Image 5 835680 842582 157418 164320 839216Image 6 827250 824643 175654 172750 825795Image 7 811893 790554 209446 188107 801519Image 8 802543 777549 222451 197457 790656Image 9 817874 784139 215861 182126 801606Image 10 804304 774794 225206 195696 790378Image 11 817725 780352 219648 182275 799755Image 12 840795 788912 211088 159205 815238Image 13 816145 787989 212011 183855 802806Image 14 798951 786152 213848 201049 793023Image 15 809012 784626 215374 190988 797600Image 16 801249 781480 218520 188751 792121Image 17 801220 781687 218318 198780 792229Image 18 782509 835148 164852 217491 807020Image 19 787041 837431 162569 212959 810816Image 20 767118 843245 156755 232882 801831
Table 3 Statistical results from the k-median clustering segmentation algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 879631 906864 93136 123069 893672Image 2 865908 901719 98281 134092 885622Image 3 833821 889051 110949 166179 862969Image 4 820844 863637 136363 179156 842695Image 5 832410 857769 142231 167590 845294Image 6 825053 842412 157588 174947 833654Image 7 811107 804281 195719 188893 807832Image 8 801857 794033 205967 198143 798186Image 9 821213 797647 202353 178787 809977Image 10 806627 791577 208423 193373 799611Image 11 820209 791588 208412 179791 806621Image 12 843809 804121 195879 156191 824514Image 13 820487 803496 196504 179513 812545Image 14 804506 794375 205625 195494 799876
Computational and Mathematical Methods in Medicine 5
Table 4 Statistical results from the PSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 875413 901196 98804 124587 890624Image 2 859612 859612 85479 85479 889689Image 3 827919 898314 101686 172081 864850Image 4 811271 887838 112162 188729 849967Image 5 826343 873995 126005 173657 850299Image 6 818996 852900 147100 181004 835581Image 7 817281 800949 199051 182719 809438Image 8 804182 800721 199279 195818 802571Image 9 822573 811450 188550 177427 817340Image 10 808521 803433 196567 191479 806182Image 11 822198 809837 190163 177802 816421Image 12 846322 815070 184930 153678 831347Image 13 826283 811153 188847 173617 819351Image 14 809338 809090 190910 190662 809226Image 15 818790 811729 188271 181210 815586Image 16 808120 814222 185778 191880 810836Image 17 808582 818136 181864 191418 812824Image 18 791387 850084 149916 208613 818114Image 19 794655 851570 148430 205345 820954Image 20 777826 853446 146554 222174 811744
Table 5 Statistical results from the IWPSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 874649 903272 96728 125351 889813Image 2 861950 898126 101874 138050 881810Image 3 829347 889622 110378 170653 861018Image 4 814285 872013 127987 185715 843584Image 5 829023 863940 136060 170977 846631Image 6 821065 845145 154855 178935 832876Image 7 821361 791744 208256 178639 807064Image 8 803274 803855 196145 196726 803544Image 9 824185 803924 196076 175815 814631Image 10 810769 795965 204035 189231 803943Image 11 822198 797401 202599 174299 812411Image 12 846322 811390 188610 152172 830334Image 13 826283 807231 173596 173596 818905Image 14 809338 808231 191769 190328 809025Image 15 818790 815899 184101 182694 816669Image 16 808120 814173 185827 191734 810896Image 17 808582 810677 189323 190166 810209Image 18 791387 849622 150378 208310 818081Image 19 794655 848219 151781 203936 820213Image 20 777684 854281 145719 222316 812033
Table 3 Continued
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 15 813002 800536 199464 186998 807248Image 16 801942 800503 199497 198058 801291Image 17 802984 803756 196244 197016 803332Image 18 786327 855226 144774 213673 817792Image 19 789322 852163 147837 210678 818439Image 20 772752 858000 142000 227248 810973
6 Computational and Mathematical Methods in Medicine
Table 6 Statistical results from the GCPSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 916158 999999 00001 83842 958079Image 2 909563 999999 00001 90437 954782Image 3 888404 999999 00001 111592 944204Image 4 872946 999999 00001 127054 936473Image 5 873583 999999 00001 126417 936792Image 6 861567 999999 00001 138433 930784Image 7 834867 999999 00001 165133 917434Image 8 831082 999999 00001 168918 915541Image 9 842907 999999 00001 157093 921453Image 10 831917 999999 00001 168083 915958Image 11 842122 999999 00001 157878 921061Image 12 857977 999999 00001 142023 928988Image 13 846397 999999 00001 153603 923198Image 14 837442 999999 00001 162558 918721Image 15 843299 999999 00001 156701 921649Image 16 838867 999999 00001 161133 919433Image 17 839061 999999 00001 160939 919531Image 18 846836 999999 00001 153164 923418Image 19 849324 999999 00001 150676 924662Image 20 839867 999999 00001 160124 919938
6500
7000
7500
8000
8500
9000
9500
True
pos
itive
rate
k-meansk-mediansPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
IWPSOGCPSO
Figure 6 Comparative results of the true positive rate value
75
80
85
90
95
100
True
neg
ativ
e rat
e
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 7 Comparative results of the true negative rate value
Computational and Mathematical Methods in Medicine 7
72
77
82
87
92
97
Acc
urac
y
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 10 Comparative results of accuracy
0
5
10
15
20
25
False
pos
itive
rate
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 8 Comparative results of the false positive rate value
0
5
10
15
20
25
False
neg
ativ
e rat
e
k-meansk-mediansPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
IWPSOGCPSO
Figure 9 Comparative results of the false negative rate value
8 Computational and Mathematical Methods in Medicine
Table 7 Statistical comparative result of accuracy
Images k-Means k-Median PSO IWPSO GCPSOImage 1 885937 893672 890624 889813 958079Image 2 882682 885622 889689 881810 954782Image 3 857501 862969 864850 861018 944204Image 4 842186 842695 849967 843584 936473Image 5 839216 845294 850299 846631 936792Image 6 825795 833654 835581 832876 930784Image 7 801519 807832 809438 807064 917434Image 8 790656 798186 802571 803544 915541Image 9 801606 809977 817340 814631 921453Image 10 790378 799611 806182 803943 915958Image 11 799755 806621 816421 812411 921061Image 12 815238 824514 831347 830334 928988Image 13 802806 812545 819351 818905 923198Image 14 793023 799876 809226 809025 918721Image 15 797600 807248 815586 816669 921649Image 16 792121 801291 810836 810896 919433Image 17 792229 803332 812824 810209 919531Image 18 807020 817792 818114 818081 923418Image 19 810816 818439 820954 820213 924662Image 20 801831 810973 811744 812033 919938
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 1
Image 2
Image 3
Image 4
Image 5
Image 6
Image 7
Image 8
Image 9
Image 10
(a)
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
(b)
Figure 11 Resultant images after preprocessing
Computational and Mathematical Methods in Medicine 9
eradicate the incidence of noise content and to improve theimage quality before an examination [4]+is part of work isknown as preprocessing In the preprocessing stage noiseremoval and contrast enhancement are two primary steps Inthe present study the performance results of medianadaptive median and average filters to isolate the presence ofspeckle noise have been compared +e coding for the samehas been implemented using MATLAB Furthermore theimage quality and visual appearance are improved byadaptive histogram equalization+e second stage of work issegmentation +is stage consists of applying five methodsnamely k-means k-median particle swarm optimization(PSO) inertia-weighted particle swarm optimization(IWPSO) and GCPSO +e tumor portion was extractedfrom the segmented results of the above-said five methodsand compared with manual extraction+e results show thatthe GCPSO-based segmentation has more accuracy than theothers Figure 1 depicts the process of operation for thepresent study
21 Median and Adaptive Median Filters +e median filterremoves the noise and retains the sharpness of the imageAccordance to the name each pixel is replaced by themedian value from the neighborhood pixels A 3 times 3 windowis used in this filter [5] +is is one of the best filters amongconventional filters which remove the speckle noise +esteps followed to construct the median filter are given inAlgorithm 1
Spatial processing to preserve the edge detail and toeliminate nonimpulsive noise by the adaptive median filterplays a vital role +e small structure in the image and edgesare retained by the adaptive median filter In the adaptivemedian filter the window size varies with respect to eachpixel
22 Average Filter +is is a simple filter which removesthe spatial noise from a digital image +e presence ofspatial noise is mainly due to the data acquisition process
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 12 Resultant images by k-means clustering
10 Computational and Mathematical Methods in Medicine
+e neighborhood mean value is measured for each andevery pixel and is replaced by the corresponding meanvalue +is process is repeated for every pixel in the image[5] All the pixels in the digital image are modified bysliding the operator over the entire range of pixels +esteps followed for the average filter are given inAlgorithm 2
23 Histogram Equalization Image enhancement is thetechnique which is used to improve the image quality Forbetter understanding and analysis it is mandatory to en-hance the contrast of medical images +e conventionalmethod used for this operation is histogram equalization Aminor adjustment on the intensity of image pixels is donein this method Each pixel is mapped to intensity pro-portional to its rank in the surrounding pixels +e stepsfollowed for histogram equalization are given in Algo-rithm 3 [6]
24 k-Means Clustering Algorithm +e simplest and con-ventional method in cluster analysis is the k-means clus-tering algorithm+is algorithm segregates the given datasetinto two or more clusters [7] +e accuracy of this methodcompletely depends on the selection of the cluster center Itis mandatory to select the optimum cluster center to get abetter result +e Euclidean distance is the general measureto segregate the dataset [8] Pixels are assigned to an indi-vidual cluster based on the Euclidean distance +e objectivefunction used in this algorithm is
J(v) 1113944C
i11113944
Ci
j1xi minus vj
1113874 11138752 (1)
where xi are the pixels vj are the cluster centers xi minus vj isthe Euclidean distance between xi and vjCi is the number ofdata points for the ith cluster and C is the number of clustercenters [9] +e steps followed for k-means clustering aregiven in Algorithm 4
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 13 Resultant images by k-median clustering
Computational and Mathematical Methods in Medicine 11
25 k-Median Clustering Algorithm +is is also a clusteringalgorithm slightly modified from the k-means algorithm Incentroid calculation instead of calculating the mean valuethe median value is considered +is algorithm significantlyreduces the error since there is no squared operation as inthe calculation of the Euclidean distance +e clustersformed by this method are more compact As an alternatethis approach uses the Lloyd-style iteration +e steps fol-lowed for k-median clustering are given in Algorithm 5 [10]
26 Particle Swarm Optimization PSO is a metaheuristicalgorithm used efficiently in medical image analysis [11] Itmimics the social behavior of the birds searching for food [12]+e fundamental idea of PSO is sharing and communicatingthe information In this approach each particle has initialposition and velocity Based on the fitness value the velocity
and position are updated +e relevant two equations in PSOto update the position and velocity are as follows [11 12]
v(t + 1) v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minus x(t)]
x(t + 1) x(t) + v(t + 1)
(2)
where r1 and r2 are the random numbers and the accel-eration coefficients c1 and c2 are two positive constants+e success of PSO relies on the fitness function +efollowing fitness function has been used for the presentstudy
maximizef 1113944n
i1
intercluster distanceintracluster distance
(3)
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 14 Resultant images by the PSO algorithm
12 Computational and Mathematical Methods in Medicine
where n is the number of clusters +e steps followed for theparticle swarm optimization are shown in Algorithm 6
27 Inertia-Weighted Particle Swarm Optimization +eexploration and exploitation in PSO are based on the inertiaweight +e basic PSO presented by Eberhart and Kennedyin 1995 has no inertia weight In 1998 Shi and Eberhartintroduced the concept of inertia weight by adding constantinertia weight +ey stated that a significant inertia weightfacilitates a global search while a small inertia weight fa-cilitates a local search [14] +is enhances the convergencerate and reduces the number of iterations Inertia weight lessthan 1 in general improves the results +e used methodimproves the convergence rate and saves the time taken andsome iterations
+e resulting velocity update equation becomes
v(t + 1) wlowast v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minusx(t)](4)
where w is the inertia weight with constant inertia weight w
07 and random inertia weight w 05 + rand()2
28 Guaranteed Convergence Particle Swarm Optimization+e GCPSO focuses on a new particle which deals with thecurrent best position in the region In this task this particleis treated as a member of the swarm and the velocity updateequation for this new particle is given as follows [15]
vφ(t + 1) xφ(t) + pbest(t) + ωvφ(t) + ρ(t)(1minus 2r) (5)
+e search ability is increased by the social part +is willimprove the random search in the area around the gbestposition +e random vector and diameter of the search area
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 15 Resultant images by IWPSO algorithm clustering
Computational and Mathematical Methods in Medicine 13
are r and ρ(t) respectively +e range of the random vectorlies between 0 and 1 +e diameter of the search area can beupdated using the following equation
ρ(t + 1)
2ρ(t) successesgt sc
115
1113874 1113875ρ(t) failuresgt fc
ρ(t) otherwise
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎩
(6)
where the terms successes and failures are defined asthe number of consecutive successes and failures re-spectively +e threshold parameters sc and fc are de-termined empirically Since it is hard to obtain a bettervalue in only a few iterations in a high-dimensional searchspace the recommended values are thus sc 15 and fc 5
On some benchmark tests the GCPSO has shown anexcellent performance of locating the minimal of a spaceafter unimodal with only a small amount of particles +esteps to be followed for the GCPSO are shown inAlgorithm 7
3 Performance Measures
Certain performance measures are used to evaluate theresults obtained from medical image segmentation +e listof performance measures used to assess the filter operation isshown in Figure 2 [16] Let If be the image after noise re-duction and I0 be the noisy image
Performance measures used for the evaluation of theresults of the segmentation algorithm are given in Figure 3[17]
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 16 Resultant images by GCPSO algorithm clustering
14 Computational and Mathematical Methods in Medicine
4 Results and Discussion
+e used methods are practically implemented usingMATLAB coding and the results were verified
In the preprocessing stage a comparison was donebetween the performance of median adaptive median andmean filters +e SSI and SMPI values are shown in Table 1and Figures 4 and 5 From the results it is evident that theadaptive median filter has accurate characteristics than themean and median filters for medical image segmentation
+e segmentation accuracy was measured using the truepositive rate true negative rate false positive rate and falsenegative rate by comparing the results from the algorithmwithmanual segmentation results +e practical results of the k-means clustering segmentation algorithm are shown in Table 2
+e practical results of the k-median clustering seg-mentation algorithm are shown in Table 3
+e practical results of the PSO-based segmentationalgorithm are shown in Table 4
+e practical results of the IWPSO segmentation algo-rithm are shown in Table 5
+e practical results of the GCPSO segmentation algo-rithm are shown in Table 6
+e graphical view of the comparison of the true positiverate true negative rate false positive rate and false negativerate for the algorithms used is shown in Figures 6ndash9 It isproved that the true positive and true negative rates are highand false positive and false negative rates are low for theGCPSO algorithm
+e comparative evaluation based on the accuracy of thesegmentation is shown in Table 7 and Figure 10 +e resultsindicate that the GCPSO-based technique has the highestaverage value of accuracy than the other methods
+e resultant images after preprocessing are shown inFigures 11(a) and 11(b)
+e resultant images after segmentation using k-meansclustering are shown in Figure 12
+e resultant images after segmentation using k-medianclustering are shown in Figure 13
+e resultant images after segmentation using the PSOalgorithm are shown in Figure 14
+e resultant images after segmentation using theIWPSO algorithm are shown in Figure 15
+e resultant images after segmentation using theGCPSO algorithm are shown in Figure 16
In an earlier research lung cancer detection was doneusing PSO genetic optimization and SVM algorithm withthe Gabor filter and produced an accuracy of 895 [18]+emethod to detect lung cancer by means of K-NN classifi-cation using the genetic algorithm produced a maximumaccuracy of 90 [19]+e comparative results with respect tothe above-said methods are shown in Table 8
+e graphical comparative analysis between the used andexisting methods is shown in Figure 17
5 Conclusion
In this study various optimization algorithms have beenevaluated to detect the tumor Medical images often need
preprocessing before being subjected to statistical analysis+e adaptive median filter has better results than medianand mean filters because the speckle suppression index andspeckle and mean preservation index values are lower for theadaptive median filter Comparing the five algorithms theaccuracy of the tumor extraction is improved in GCPSOwith the highest accuracy of 958079 and it obtained above90 of precision in all the 20 images It is more accuratewhen compared to the previous method which had an ac-curacy of 90 in 4 out of 10 datasets only In future studiesthe use of more number of optimization algorithms will beincluded to improve the accuracy
Data Availability
+eCTimages data used to support the findings of this studyhave been deposited in the LungCT-Diagnosis repository(doiorg107937K9TCIA2015A6V7JIWX)
Conflicts of Interest
+e authors declare that there are no conflicts of interestregarding the publication of this paper
References
[1] A A Brindha S Indirani and A Srinivasan ldquoLung cancerdetection using SVM algorithm and optimization techniquesrdquoJournal of Chemical and Pharmaceutical Sciences vol 9 no 42016
[2] M Kurkure and A +akare ldquoIntroducing automated systemfor lung cancer detection using Evolutionary ApproachrdquoInternational Journal of Engineering and Computer Sciencevol 5 no 5 pp 16736ndash16739 2016
[3] B Rani A K Goel and R Kaur ldquoA modified approach forlung cancer detection using bacterial forging optimization
868890929496
PSO GA and SVM algorithm
K-NNclassification
using GA
Proposed GCPSOmethod
8950 90
9581
Acc
urac
y
Various algorithms
Comparison between existing and projected methods
Figure 17 Graphical view of accuracy
Table 8 Comparative analysis of accuracy of the projected methodwith various methods
Various methods Accuracy ()PSO GA and SVM algorithm [18] 8950K-NN classification using GA [19] 9000Projected GCPSO method 9581
Computational and Mathematical Methods in Medicine 15
algorithmrdquo International Journal of Scientific Research En-gineering and Technology vol 5 no 1 2016
[4] N Panpaliya N Tadas S Bobade R Aglawe andA Gudadhe ldquoA survey on early detection and prediction oflung cancerrdquo International Journal of Computer Science andMobile Computing vol 4 no 1 pp 175ndash184 2015
[5] G Gupta ldquoAlgorithm for image processing using improvedmedian filter and comparison of mean median and improvedmedian filterrdquo International Journal of Soft Computing andEngineering vol 1 no 5 2011
[6] Tutorial point with digital image processing httpwwwtutorialspointcomdiphistogram_equalizationhtm
[7] K Venkatalakshmi and S M Shalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and K-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Melbourne VIC Australia December2005
[8] K Venkatalakshmi and S M Shalinie ldquoMultispectral imageclassification using modified k-means clusteringrdquo In-ternational Journal on Neural and Mass-Parallel Computingand Information Systems vol 17 no 2 pp 113ndash120 2007
[9] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercy Shalinie ldquoMultispectral image clustering using en-hanced genetic k-means algorithmrdquo Information TechnologyJournal vol 6 no 4 pp 554ndash560 2007
[10] P I Dalatu ldquoTime complexity of k-means and k-mediansclustering algorithms in outliers detectionrdquo Global Journal ofPure and Applied Mathematics vol 12 no 5 pp 4405ndash44182016
[11] K Senthilkumar K Venkatalakshmi K Karthikeyan andP Kathirkamasundari ldquoAn efficient method for segmentingdigital image using a hybrid model of particle swarm opti-mization and artificial bee colony algorithmrdquo InternationalJournal of Applied Engineering Research vol 10 pp 444ndash4492015
[12] K Venkatalakshmi and S Mercyshalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and k-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Bangalore India December 2005
[13] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercyshalinie ldquoA customized particle swarm optimizationfor classification of multispectral imagery based on featurefusionrdquo International Arab Journal of Information Technologyvol 5 no 4 pp 327ndash333 2008
[14] J C Bansal and P K Singh ldquoInertia weight strategies inparticle swarm optimizationrdquo in Proceedings of IEEE WorldCongress on Nature and Biologically Inspired Computing pp640ndash647 Salamanca Spain October 2011
[15] P K Patel V Sharma and K Gupta ldquoGuaranteed conver-gence particle swarm optimization using Personal Bestrdquo In-ternational Journal of Computer Applications vol 73 no 7pp 6ndash10 2013
[16] X Wang L Ge and X Li ldquoEvaluation of filters for Envi-satAsar speckle suppression in pasture areardquo ISPRS Annals ofthe Photogrammetry Remote Sensing and Spatial InformationSciences vol 1ndash7 pp 341ndash346 2012
[17] H Nagaveena D Devaraj and S C Prasanna KumarldquoVessels segmentation in diabetic retinopathy by adaptivemedian thresholdingrdquo International Journal of Science andTechnology vol 1 no 1 pp 17ndash22 2013
[18] A Asuntha N Singh and A Srinivasan ldquoPSO genetic op-timization and SVM algorithm used for lung cancer
detectionrdquo Journal of Chemical and Pharmaceutical Researchvol 8 no 6 pp 351ndash359 2016
[19] P Bhuvaneswari and A Brintha+erese ldquoDetection of cancerin lung with K-NN classification using genetic algorithmrdquoInternational Conference on Nanomaterials and Technologiesvol 10 pp 433ndash440 2014
16 Computational and Mathematical Methods in Medicine
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom
True positive (tp)Pixels correctly segmented
as foreground
True negative (tn)Pixels correctly detected as
background
False positive (fp)Pixels falsely segmented as
foreground
False negative (fn)Pixels falsely detected as
background
AccuracyA degree of measure to
state the correctness of aprocess
tp + tnAccuracy =
tp + tn + fp + fn
tpr =tp
tp + fn
fnr =fn
fn + tpfpr =
fpfp + tn
tnr =tn
tn + fp
Figure 3 Performance measures for the medical image segmentation
Table 1 SSI and SMPI values of input images
Sample imagesSSI SMPI
Mean filter Median filter Adaptive median filter Mean filter Median filter Adaptive median filterImage 1 09621 08208 08086 09857 09788 09638Image 2 09658 08232 08087 09895 0959 09452Image 3 09588 08209 08091 09883 09799 09696Image 4 09671 08080 07937 09958 09836 09703Image 5 09705 08220 08078 09851 09833 09706Image 6 09708 08218 07900 09948 09775 09457Image 7 09660 08202 08067 09979 09608 09464Image 8 09640 08265 08154 09922 09622 09493Image 9 09638 08272 08141 09990 09716 09576Image 10 09644 08238 08112 09944 09804 09659Image 11 09639 08231 08122 09765 09788 09643Image 12 09642 08289 08152 10012 09826 09721Image 13 09648 08239 08135 09920 09782 09674Image 14 09564 08242 08098 09888 09767 09648Image 15 09573 08208 08084 10005 09785 09636Image 16 09631 08242 08095 09912 09755 09613Image 17 09919 08239 08352 09722 09770 09882Image 18 09912 07983 07857 10003 09808 09696Image 19 09921 08020 07884 10037 09838 09706Image 20 09939 08085 07690 09968 09741 09432
075
08
085
09
095
1
105
SSI v
alue
Sample images
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Mean filterMedian filterAdaptive median filter
Figure 4 Comparative results of SSI values
4 Computational and Mathematical Methods in Medicine
Sample images
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Mean filterMedian filterAdaptive median filter
093094095096097098099
1101
SMPI
val
ue
Figure 5 Comparative results of SMPI values
Table 2 Statistical results from the k-means algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 878783 891554 108446 121217 885937Image 2 866527 895874 104126 133473 882682Image 3 838975 873900 126100 161025 857501Image 4 826502 857011 142989 173498 842186Image 5 835680 842582 157418 164320 839216Image 6 827250 824643 175654 172750 825795Image 7 811893 790554 209446 188107 801519Image 8 802543 777549 222451 197457 790656Image 9 817874 784139 215861 182126 801606Image 10 804304 774794 225206 195696 790378Image 11 817725 780352 219648 182275 799755Image 12 840795 788912 211088 159205 815238Image 13 816145 787989 212011 183855 802806Image 14 798951 786152 213848 201049 793023Image 15 809012 784626 215374 190988 797600Image 16 801249 781480 218520 188751 792121Image 17 801220 781687 218318 198780 792229Image 18 782509 835148 164852 217491 807020Image 19 787041 837431 162569 212959 810816Image 20 767118 843245 156755 232882 801831
Table 3 Statistical results from the k-median clustering segmentation algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 879631 906864 93136 123069 893672Image 2 865908 901719 98281 134092 885622Image 3 833821 889051 110949 166179 862969Image 4 820844 863637 136363 179156 842695Image 5 832410 857769 142231 167590 845294Image 6 825053 842412 157588 174947 833654Image 7 811107 804281 195719 188893 807832Image 8 801857 794033 205967 198143 798186Image 9 821213 797647 202353 178787 809977Image 10 806627 791577 208423 193373 799611Image 11 820209 791588 208412 179791 806621Image 12 843809 804121 195879 156191 824514Image 13 820487 803496 196504 179513 812545Image 14 804506 794375 205625 195494 799876
Computational and Mathematical Methods in Medicine 5
Table 4 Statistical results from the PSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 875413 901196 98804 124587 890624Image 2 859612 859612 85479 85479 889689Image 3 827919 898314 101686 172081 864850Image 4 811271 887838 112162 188729 849967Image 5 826343 873995 126005 173657 850299Image 6 818996 852900 147100 181004 835581Image 7 817281 800949 199051 182719 809438Image 8 804182 800721 199279 195818 802571Image 9 822573 811450 188550 177427 817340Image 10 808521 803433 196567 191479 806182Image 11 822198 809837 190163 177802 816421Image 12 846322 815070 184930 153678 831347Image 13 826283 811153 188847 173617 819351Image 14 809338 809090 190910 190662 809226Image 15 818790 811729 188271 181210 815586Image 16 808120 814222 185778 191880 810836Image 17 808582 818136 181864 191418 812824Image 18 791387 850084 149916 208613 818114Image 19 794655 851570 148430 205345 820954Image 20 777826 853446 146554 222174 811744
Table 5 Statistical results from the IWPSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 874649 903272 96728 125351 889813Image 2 861950 898126 101874 138050 881810Image 3 829347 889622 110378 170653 861018Image 4 814285 872013 127987 185715 843584Image 5 829023 863940 136060 170977 846631Image 6 821065 845145 154855 178935 832876Image 7 821361 791744 208256 178639 807064Image 8 803274 803855 196145 196726 803544Image 9 824185 803924 196076 175815 814631Image 10 810769 795965 204035 189231 803943Image 11 822198 797401 202599 174299 812411Image 12 846322 811390 188610 152172 830334Image 13 826283 807231 173596 173596 818905Image 14 809338 808231 191769 190328 809025Image 15 818790 815899 184101 182694 816669Image 16 808120 814173 185827 191734 810896Image 17 808582 810677 189323 190166 810209Image 18 791387 849622 150378 208310 818081Image 19 794655 848219 151781 203936 820213Image 20 777684 854281 145719 222316 812033
Table 3 Continued
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 15 813002 800536 199464 186998 807248Image 16 801942 800503 199497 198058 801291Image 17 802984 803756 196244 197016 803332Image 18 786327 855226 144774 213673 817792Image 19 789322 852163 147837 210678 818439Image 20 772752 858000 142000 227248 810973
6 Computational and Mathematical Methods in Medicine
Table 6 Statistical results from the GCPSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 916158 999999 00001 83842 958079Image 2 909563 999999 00001 90437 954782Image 3 888404 999999 00001 111592 944204Image 4 872946 999999 00001 127054 936473Image 5 873583 999999 00001 126417 936792Image 6 861567 999999 00001 138433 930784Image 7 834867 999999 00001 165133 917434Image 8 831082 999999 00001 168918 915541Image 9 842907 999999 00001 157093 921453Image 10 831917 999999 00001 168083 915958Image 11 842122 999999 00001 157878 921061Image 12 857977 999999 00001 142023 928988Image 13 846397 999999 00001 153603 923198Image 14 837442 999999 00001 162558 918721Image 15 843299 999999 00001 156701 921649Image 16 838867 999999 00001 161133 919433Image 17 839061 999999 00001 160939 919531Image 18 846836 999999 00001 153164 923418Image 19 849324 999999 00001 150676 924662Image 20 839867 999999 00001 160124 919938
6500
7000
7500
8000
8500
9000
9500
True
pos
itive
rate
k-meansk-mediansPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
IWPSOGCPSO
Figure 6 Comparative results of the true positive rate value
75
80
85
90
95
100
True
neg
ativ
e rat
e
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 7 Comparative results of the true negative rate value
Computational and Mathematical Methods in Medicine 7
72
77
82
87
92
97
Acc
urac
y
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 10 Comparative results of accuracy
0
5
10
15
20
25
False
pos
itive
rate
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 8 Comparative results of the false positive rate value
0
5
10
15
20
25
False
neg
ativ
e rat
e
k-meansk-mediansPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
IWPSOGCPSO
Figure 9 Comparative results of the false negative rate value
8 Computational and Mathematical Methods in Medicine
Table 7 Statistical comparative result of accuracy
Images k-Means k-Median PSO IWPSO GCPSOImage 1 885937 893672 890624 889813 958079Image 2 882682 885622 889689 881810 954782Image 3 857501 862969 864850 861018 944204Image 4 842186 842695 849967 843584 936473Image 5 839216 845294 850299 846631 936792Image 6 825795 833654 835581 832876 930784Image 7 801519 807832 809438 807064 917434Image 8 790656 798186 802571 803544 915541Image 9 801606 809977 817340 814631 921453Image 10 790378 799611 806182 803943 915958Image 11 799755 806621 816421 812411 921061Image 12 815238 824514 831347 830334 928988Image 13 802806 812545 819351 818905 923198Image 14 793023 799876 809226 809025 918721Image 15 797600 807248 815586 816669 921649Image 16 792121 801291 810836 810896 919433Image 17 792229 803332 812824 810209 919531Image 18 807020 817792 818114 818081 923418Image 19 810816 818439 820954 820213 924662Image 20 801831 810973 811744 812033 919938
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 1
Image 2
Image 3
Image 4
Image 5
Image 6
Image 7
Image 8
Image 9
Image 10
(a)
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
(b)
Figure 11 Resultant images after preprocessing
Computational and Mathematical Methods in Medicine 9
eradicate the incidence of noise content and to improve theimage quality before an examination [4]+is part of work isknown as preprocessing In the preprocessing stage noiseremoval and contrast enhancement are two primary steps Inthe present study the performance results of medianadaptive median and average filters to isolate the presence ofspeckle noise have been compared +e coding for the samehas been implemented using MATLAB Furthermore theimage quality and visual appearance are improved byadaptive histogram equalization+e second stage of work issegmentation +is stage consists of applying five methodsnamely k-means k-median particle swarm optimization(PSO) inertia-weighted particle swarm optimization(IWPSO) and GCPSO +e tumor portion was extractedfrom the segmented results of the above-said five methodsand compared with manual extraction+e results show thatthe GCPSO-based segmentation has more accuracy than theothers Figure 1 depicts the process of operation for thepresent study
21 Median and Adaptive Median Filters +e median filterremoves the noise and retains the sharpness of the imageAccordance to the name each pixel is replaced by themedian value from the neighborhood pixels A 3 times 3 windowis used in this filter [5] +is is one of the best filters amongconventional filters which remove the speckle noise +esteps followed to construct the median filter are given inAlgorithm 1
Spatial processing to preserve the edge detail and toeliminate nonimpulsive noise by the adaptive median filterplays a vital role +e small structure in the image and edgesare retained by the adaptive median filter In the adaptivemedian filter the window size varies with respect to eachpixel
22 Average Filter +is is a simple filter which removesthe spatial noise from a digital image +e presence ofspatial noise is mainly due to the data acquisition process
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 12 Resultant images by k-means clustering
10 Computational and Mathematical Methods in Medicine
+e neighborhood mean value is measured for each andevery pixel and is replaced by the corresponding meanvalue +is process is repeated for every pixel in the image[5] All the pixels in the digital image are modified bysliding the operator over the entire range of pixels +esteps followed for the average filter are given inAlgorithm 2
23 Histogram Equalization Image enhancement is thetechnique which is used to improve the image quality Forbetter understanding and analysis it is mandatory to en-hance the contrast of medical images +e conventionalmethod used for this operation is histogram equalization Aminor adjustment on the intensity of image pixels is donein this method Each pixel is mapped to intensity pro-portional to its rank in the surrounding pixels +e stepsfollowed for histogram equalization are given in Algo-rithm 3 [6]
24 k-Means Clustering Algorithm +e simplest and con-ventional method in cluster analysis is the k-means clus-tering algorithm+is algorithm segregates the given datasetinto two or more clusters [7] +e accuracy of this methodcompletely depends on the selection of the cluster center Itis mandatory to select the optimum cluster center to get abetter result +e Euclidean distance is the general measureto segregate the dataset [8] Pixels are assigned to an indi-vidual cluster based on the Euclidean distance +e objectivefunction used in this algorithm is
J(v) 1113944C
i11113944
Ci
j1xi minus vj
1113874 11138752 (1)
where xi are the pixels vj are the cluster centers xi minus vj isthe Euclidean distance between xi and vjCi is the number ofdata points for the ith cluster and C is the number of clustercenters [9] +e steps followed for k-means clustering aregiven in Algorithm 4
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 13 Resultant images by k-median clustering
Computational and Mathematical Methods in Medicine 11
25 k-Median Clustering Algorithm +is is also a clusteringalgorithm slightly modified from the k-means algorithm Incentroid calculation instead of calculating the mean valuethe median value is considered +is algorithm significantlyreduces the error since there is no squared operation as inthe calculation of the Euclidean distance +e clustersformed by this method are more compact As an alternatethis approach uses the Lloyd-style iteration +e steps fol-lowed for k-median clustering are given in Algorithm 5 [10]
26 Particle Swarm Optimization PSO is a metaheuristicalgorithm used efficiently in medical image analysis [11] Itmimics the social behavior of the birds searching for food [12]+e fundamental idea of PSO is sharing and communicatingthe information In this approach each particle has initialposition and velocity Based on the fitness value the velocity
and position are updated +e relevant two equations in PSOto update the position and velocity are as follows [11 12]
v(t + 1) v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minus x(t)]
x(t + 1) x(t) + v(t + 1)
(2)
where r1 and r2 are the random numbers and the accel-eration coefficients c1 and c2 are two positive constants+e success of PSO relies on the fitness function +efollowing fitness function has been used for the presentstudy
maximizef 1113944n
i1
intercluster distanceintracluster distance
(3)
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 14 Resultant images by the PSO algorithm
12 Computational and Mathematical Methods in Medicine
where n is the number of clusters +e steps followed for theparticle swarm optimization are shown in Algorithm 6
27 Inertia-Weighted Particle Swarm Optimization +eexploration and exploitation in PSO are based on the inertiaweight +e basic PSO presented by Eberhart and Kennedyin 1995 has no inertia weight In 1998 Shi and Eberhartintroduced the concept of inertia weight by adding constantinertia weight +ey stated that a significant inertia weightfacilitates a global search while a small inertia weight fa-cilitates a local search [14] +is enhances the convergencerate and reduces the number of iterations Inertia weight lessthan 1 in general improves the results +e used methodimproves the convergence rate and saves the time taken andsome iterations
+e resulting velocity update equation becomes
v(t + 1) wlowast v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minusx(t)](4)
where w is the inertia weight with constant inertia weight w
07 and random inertia weight w 05 + rand()2
28 Guaranteed Convergence Particle Swarm Optimization+e GCPSO focuses on a new particle which deals with thecurrent best position in the region In this task this particleis treated as a member of the swarm and the velocity updateequation for this new particle is given as follows [15]
vφ(t + 1) xφ(t) + pbest(t) + ωvφ(t) + ρ(t)(1minus 2r) (5)
+e search ability is increased by the social part +is willimprove the random search in the area around the gbestposition +e random vector and diameter of the search area
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 15 Resultant images by IWPSO algorithm clustering
Computational and Mathematical Methods in Medicine 13
are r and ρ(t) respectively +e range of the random vectorlies between 0 and 1 +e diameter of the search area can beupdated using the following equation
ρ(t + 1)
2ρ(t) successesgt sc
115
1113874 1113875ρ(t) failuresgt fc
ρ(t) otherwise
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎩
(6)
where the terms successes and failures are defined asthe number of consecutive successes and failures re-spectively +e threshold parameters sc and fc are de-termined empirically Since it is hard to obtain a bettervalue in only a few iterations in a high-dimensional searchspace the recommended values are thus sc 15 and fc 5
On some benchmark tests the GCPSO has shown anexcellent performance of locating the minimal of a spaceafter unimodal with only a small amount of particles +esteps to be followed for the GCPSO are shown inAlgorithm 7
3 Performance Measures
Certain performance measures are used to evaluate theresults obtained from medical image segmentation +e listof performance measures used to assess the filter operation isshown in Figure 2 [16] Let If be the image after noise re-duction and I0 be the noisy image
Performance measures used for the evaluation of theresults of the segmentation algorithm are given in Figure 3[17]
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 16 Resultant images by GCPSO algorithm clustering
14 Computational and Mathematical Methods in Medicine
4 Results and Discussion
+e used methods are practically implemented usingMATLAB coding and the results were verified
In the preprocessing stage a comparison was donebetween the performance of median adaptive median andmean filters +e SSI and SMPI values are shown in Table 1and Figures 4 and 5 From the results it is evident that theadaptive median filter has accurate characteristics than themean and median filters for medical image segmentation
+e segmentation accuracy was measured using the truepositive rate true negative rate false positive rate and falsenegative rate by comparing the results from the algorithmwithmanual segmentation results +e practical results of the k-means clustering segmentation algorithm are shown in Table 2
+e practical results of the k-median clustering seg-mentation algorithm are shown in Table 3
+e practical results of the PSO-based segmentationalgorithm are shown in Table 4
+e practical results of the IWPSO segmentation algo-rithm are shown in Table 5
+e practical results of the GCPSO segmentation algo-rithm are shown in Table 6
+e graphical view of the comparison of the true positiverate true negative rate false positive rate and false negativerate for the algorithms used is shown in Figures 6ndash9 It isproved that the true positive and true negative rates are highand false positive and false negative rates are low for theGCPSO algorithm
+e comparative evaluation based on the accuracy of thesegmentation is shown in Table 7 and Figure 10 +e resultsindicate that the GCPSO-based technique has the highestaverage value of accuracy than the other methods
+e resultant images after preprocessing are shown inFigures 11(a) and 11(b)
+e resultant images after segmentation using k-meansclustering are shown in Figure 12
+e resultant images after segmentation using k-medianclustering are shown in Figure 13
+e resultant images after segmentation using the PSOalgorithm are shown in Figure 14
+e resultant images after segmentation using theIWPSO algorithm are shown in Figure 15
+e resultant images after segmentation using theGCPSO algorithm are shown in Figure 16
In an earlier research lung cancer detection was doneusing PSO genetic optimization and SVM algorithm withthe Gabor filter and produced an accuracy of 895 [18]+emethod to detect lung cancer by means of K-NN classifi-cation using the genetic algorithm produced a maximumaccuracy of 90 [19]+e comparative results with respect tothe above-said methods are shown in Table 8
+e graphical comparative analysis between the used andexisting methods is shown in Figure 17
5 Conclusion
In this study various optimization algorithms have beenevaluated to detect the tumor Medical images often need
preprocessing before being subjected to statistical analysis+e adaptive median filter has better results than medianand mean filters because the speckle suppression index andspeckle and mean preservation index values are lower for theadaptive median filter Comparing the five algorithms theaccuracy of the tumor extraction is improved in GCPSOwith the highest accuracy of 958079 and it obtained above90 of precision in all the 20 images It is more accuratewhen compared to the previous method which had an ac-curacy of 90 in 4 out of 10 datasets only In future studiesthe use of more number of optimization algorithms will beincluded to improve the accuracy
Data Availability
+eCTimages data used to support the findings of this studyhave been deposited in the LungCT-Diagnosis repository(doiorg107937K9TCIA2015A6V7JIWX)
Conflicts of Interest
+e authors declare that there are no conflicts of interestregarding the publication of this paper
References
[1] A A Brindha S Indirani and A Srinivasan ldquoLung cancerdetection using SVM algorithm and optimization techniquesrdquoJournal of Chemical and Pharmaceutical Sciences vol 9 no 42016
[2] M Kurkure and A +akare ldquoIntroducing automated systemfor lung cancer detection using Evolutionary ApproachrdquoInternational Journal of Engineering and Computer Sciencevol 5 no 5 pp 16736ndash16739 2016
[3] B Rani A K Goel and R Kaur ldquoA modified approach forlung cancer detection using bacterial forging optimization
868890929496
PSO GA and SVM algorithm
K-NNclassification
using GA
Proposed GCPSOmethod
8950 90
9581
Acc
urac
y
Various algorithms
Comparison between existing and projected methods
Figure 17 Graphical view of accuracy
Table 8 Comparative analysis of accuracy of the projected methodwith various methods
Various methods Accuracy ()PSO GA and SVM algorithm [18] 8950K-NN classification using GA [19] 9000Projected GCPSO method 9581
Computational and Mathematical Methods in Medicine 15
algorithmrdquo International Journal of Scientific Research En-gineering and Technology vol 5 no 1 2016
[4] N Panpaliya N Tadas S Bobade R Aglawe andA Gudadhe ldquoA survey on early detection and prediction oflung cancerrdquo International Journal of Computer Science andMobile Computing vol 4 no 1 pp 175ndash184 2015
[5] G Gupta ldquoAlgorithm for image processing using improvedmedian filter and comparison of mean median and improvedmedian filterrdquo International Journal of Soft Computing andEngineering vol 1 no 5 2011
[6] Tutorial point with digital image processing httpwwwtutorialspointcomdiphistogram_equalizationhtm
[7] K Venkatalakshmi and S M Shalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and K-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Melbourne VIC Australia December2005
[8] K Venkatalakshmi and S M Shalinie ldquoMultispectral imageclassification using modified k-means clusteringrdquo In-ternational Journal on Neural and Mass-Parallel Computingand Information Systems vol 17 no 2 pp 113ndash120 2007
[9] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercy Shalinie ldquoMultispectral image clustering using en-hanced genetic k-means algorithmrdquo Information TechnologyJournal vol 6 no 4 pp 554ndash560 2007
[10] P I Dalatu ldquoTime complexity of k-means and k-mediansclustering algorithms in outliers detectionrdquo Global Journal ofPure and Applied Mathematics vol 12 no 5 pp 4405ndash44182016
[11] K Senthilkumar K Venkatalakshmi K Karthikeyan andP Kathirkamasundari ldquoAn efficient method for segmentingdigital image using a hybrid model of particle swarm opti-mization and artificial bee colony algorithmrdquo InternationalJournal of Applied Engineering Research vol 10 pp 444ndash4492015
[12] K Venkatalakshmi and S Mercyshalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and k-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Bangalore India December 2005
[13] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercyshalinie ldquoA customized particle swarm optimizationfor classification of multispectral imagery based on featurefusionrdquo International Arab Journal of Information Technologyvol 5 no 4 pp 327ndash333 2008
[14] J C Bansal and P K Singh ldquoInertia weight strategies inparticle swarm optimizationrdquo in Proceedings of IEEE WorldCongress on Nature and Biologically Inspired Computing pp640ndash647 Salamanca Spain October 2011
[15] P K Patel V Sharma and K Gupta ldquoGuaranteed conver-gence particle swarm optimization using Personal Bestrdquo In-ternational Journal of Computer Applications vol 73 no 7pp 6ndash10 2013
[16] X Wang L Ge and X Li ldquoEvaluation of filters for Envi-satAsar speckle suppression in pasture areardquo ISPRS Annals ofthe Photogrammetry Remote Sensing and Spatial InformationSciences vol 1ndash7 pp 341ndash346 2012
[17] H Nagaveena D Devaraj and S C Prasanna KumarldquoVessels segmentation in diabetic retinopathy by adaptivemedian thresholdingrdquo International Journal of Science andTechnology vol 1 no 1 pp 17ndash22 2013
[18] A Asuntha N Singh and A Srinivasan ldquoPSO genetic op-timization and SVM algorithm used for lung cancer
detectionrdquo Journal of Chemical and Pharmaceutical Researchvol 8 no 6 pp 351ndash359 2016
[19] P Bhuvaneswari and A Brintha+erese ldquoDetection of cancerin lung with K-NN classification using genetic algorithmrdquoInternational Conference on Nanomaterials and Technologiesvol 10 pp 433ndash440 2014
16 Computational and Mathematical Methods in Medicine
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom
Sample images
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Mean filterMedian filterAdaptive median filter
093094095096097098099
1101
SMPI
val
ue
Figure 5 Comparative results of SMPI values
Table 2 Statistical results from the k-means algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 878783 891554 108446 121217 885937Image 2 866527 895874 104126 133473 882682Image 3 838975 873900 126100 161025 857501Image 4 826502 857011 142989 173498 842186Image 5 835680 842582 157418 164320 839216Image 6 827250 824643 175654 172750 825795Image 7 811893 790554 209446 188107 801519Image 8 802543 777549 222451 197457 790656Image 9 817874 784139 215861 182126 801606Image 10 804304 774794 225206 195696 790378Image 11 817725 780352 219648 182275 799755Image 12 840795 788912 211088 159205 815238Image 13 816145 787989 212011 183855 802806Image 14 798951 786152 213848 201049 793023Image 15 809012 784626 215374 190988 797600Image 16 801249 781480 218520 188751 792121Image 17 801220 781687 218318 198780 792229Image 18 782509 835148 164852 217491 807020Image 19 787041 837431 162569 212959 810816Image 20 767118 843245 156755 232882 801831
Table 3 Statistical results from the k-median clustering segmentation algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 879631 906864 93136 123069 893672Image 2 865908 901719 98281 134092 885622Image 3 833821 889051 110949 166179 862969Image 4 820844 863637 136363 179156 842695Image 5 832410 857769 142231 167590 845294Image 6 825053 842412 157588 174947 833654Image 7 811107 804281 195719 188893 807832Image 8 801857 794033 205967 198143 798186Image 9 821213 797647 202353 178787 809977Image 10 806627 791577 208423 193373 799611Image 11 820209 791588 208412 179791 806621Image 12 843809 804121 195879 156191 824514Image 13 820487 803496 196504 179513 812545Image 14 804506 794375 205625 195494 799876
Computational and Mathematical Methods in Medicine 5
Table 4 Statistical results from the PSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 875413 901196 98804 124587 890624Image 2 859612 859612 85479 85479 889689Image 3 827919 898314 101686 172081 864850Image 4 811271 887838 112162 188729 849967Image 5 826343 873995 126005 173657 850299Image 6 818996 852900 147100 181004 835581Image 7 817281 800949 199051 182719 809438Image 8 804182 800721 199279 195818 802571Image 9 822573 811450 188550 177427 817340Image 10 808521 803433 196567 191479 806182Image 11 822198 809837 190163 177802 816421Image 12 846322 815070 184930 153678 831347Image 13 826283 811153 188847 173617 819351Image 14 809338 809090 190910 190662 809226Image 15 818790 811729 188271 181210 815586Image 16 808120 814222 185778 191880 810836Image 17 808582 818136 181864 191418 812824Image 18 791387 850084 149916 208613 818114Image 19 794655 851570 148430 205345 820954Image 20 777826 853446 146554 222174 811744
Table 5 Statistical results from the IWPSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 874649 903272 96728 125351 889813Image 2 861950 898126 101874 138050 881810Image 3 829347 889622 110378 170653 861018Image 4 814285 872013 127987 185715 843584Image 5 829023 863940 136060 170977 846631Image 6 821065 845145 154855 178935 832876Image 7 821361 791744 208256 178639 807064Image 8 803274 803855 196145 196726 803544Image 9 824185 803924 196076 175815 814631Image 10 810769 795965 204035 189231 803943Image 11 822198 797401 202599 174299 812411Image 12 846322 811390 188610 152172 830334Image 13 826283 807231 173596 173596 818905Image 14 809338 808231 191769 190328 809025Image 15 818790 815899 184101 182694 816669Image 16 808120 814173 185827 191734 810896Image 17 808582 810677 189323 190166 810209Image 18 791387 849622 150378 208310 818081Image 19 794655 848219 151781 203936 820213Image 20 777684 854281 145719 222316 812033
Table 3 Continued
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 15 813002 800536 199464 186998 807248Image 16 801942 800503 199497 198058 801291Image 17 802984 803756 196244 197016 803332Image 18 786327 855226 144774 213673 817792Image 19 789322 852163 147837 210678 818439Image 20 772752 858000 142000 227248 810973
6 Computational and Mathematical Methods in Medicine
Table 6 Statistical results from the GCPSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 916158 999999 00001 83842 958079Image 2 909563 999999 00001 90437 954782Image 3 888404 999999 00001 111592 944204Image 4 872946 999999 00001 127054 936473Image 5 873583 999999 00001 126417 936792Image 6 861567 999999 00001 138433 930784Image 7 834867 999999 00001 165133 917434Image 8 831082 999999 00001 168918 915541Image 9 842907 999999 00001 157093 921453Image 10 831917 999999 00001 168083 915958Image 11 842122 999999 00001 157878 921061Image 12 857977 999999 00001 142023 928988Image 13 846397 999999 00001 153603 923198Image 14 837442 999999 00001 162558 918721Image 15 843299 999999 00001 156701 921649Image 16 838867 999999 00001 161133 919433Image 17 839061 999999 00001 160939 919531Image 18 846836 999999 00001 153164 923418Image 19 849324 999999 00001 150676 924662Image 20 839867 999999 00001 160124 919938
6500
7000
7500
8000
8500
9000
9500
True
pos
itive
rate
k-meansk-mediansPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
IWPSOGCPSO
Figure 6 Comparative results of the true positive rate value
75
80
85
90
95
100
True
neg
ativ
e rat
e
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 7 Comparative results of the true negative rate value
Computational and Mathematical Methods in Medicine 7
72
77
82
87
92
97
Acc
urac
y
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 10 Comparative results of accuracy
0
5
10
15
20
25
False
pos
itive
rate
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 8 Comparative results of the false positive rate value
0
5
10
15
20
25
False
neg
ativ
e rat
e
k-meansk-mediansPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
IWPSOGCPSO
Figure 9 Comparative results of the false negative rate value
8 Computational and Mathematical Methods in Medicine
Table 7 Statistical comparative result of accuracy
Images k-Means k-Median PSO IWPSO GCPSOImage 1 885937 893672 890624 889813 958079Image 2 882682 885622 889689 881810 954782Image 3 857501 862969 864850 861018 944204Image 4 842186 842695 849967 843584 936473Image 5 839216 845294 850299 846631 936792Image 6 825795 833654 835581 832876 930784Image 7 801519 807832 809438 807064 917434Image 8 790656 798186 802571 803544 915541Image 9 801606 809977 817340 814631 921453Image 10 790378 799611 806182 803943 915958Image 11 799755 806621 816421 812411 921061Image 12 815238 824514 831347 830334 928988Image 13 802806 812545 819351 818905 923198Image 14 793023 799876 809226 809025 918721Image 15 797600 807248 815586 816669 921649Image 16 792121 801291 810836 810896 919433Image 17 792229 803332 812824 810209 919531Image 18 807020 817792 818114 818081 923418Image 19 810816 818439 820954 820213 924662Image 20 801831 810973 811744 812033 919938
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 1
Image 2
Image 3
Image 4
Image 5
Image 6
Image 7
Image 8
Image 9
Image 10
(a)
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
(b)
Figure 11 Resultant images after preprocessing
Computational and Mathematical Methods in Medicine 9
eradicate the incidence of noise content and to improve theimage quality before an examination [4]+is part of work isknown as preprocessing In the preprocessing stage noiseremoval and contrast enhancement are two primary steps Inthe present study the performance results of medianadaptive median and average filters to isolate the presence ofspeckle noise have been compared +e coding for the samehas been implemented using MATLAB Furthermore theimage quality and visual appearance are improved byadaptive histogram equalization+e second stage of work issegmentation +is stage consists of applying five methodsnamely k-means k-median particle swarm optimization(PSO) inertia-weighted particle swarm optimization(IWPSO) and GCPSO +e tumor portion was extractedfrom the segmented results of the above-said five methodsand compared with manual extraction+e results show thatthe GCPSO-based segmentation has more accuracy than theothers Figure 1 depicts the process of operation for thepresent study
21 Median and Adaptive Median Filters +e median filterremoves the noise and retains the sharpness of the imageAccordance to the name each pixel is replaced by themedian value from the neighborhood pixels A 3 times 3 windowis used in this filter [5] +is is one of the best filters amongconventional filters which remove the speckle noise +esteps followed to construct the median filter are given inAlgorithm 1
Spatial processing to preserve the edge detail and toeliminate nonimpulsive noise by the adaptive median filterplays a vital role +e small structure in the image and edgesare retained by the adaptive median filter In the adaptivemedian filter the window size varies with respect to eachpixel
22 Average Filter +is is a simple filter which removesthe spatial noise from a digital image +e presence ofspatial noise is mainly due to the data acquisition process
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 12 Resultant images by k-means clustering
10 Computational and Mathematical Methods in Medicine
+e neighborhood mean value is measured for each andevery pixel and is replaced by the corresponding meanvalue +is process is repeated for every pixel in the image[5] All the pixels in the digital image are modified bysliding the operator over the entire range of pixels +esteps followed for the average filter are given inAlgorithm 2
23 Histogram Equalization Image enhancement is thetechnique which is used to improve the image quality Forbetter understanding and analysis it is mandatory to en-hance the contrast of medical images +e conventionalmethod used for this operation is histogram equalization Aminor adjustment on the intensity of image pixels is donein this method Each pixel is mapped to intensity pro-portional to its rank in the surrounding pixels +e stepsfollowed for histogram equalization are given in Algo-rithm 3 [6]
24 k-Means Clustering Algorithm +e simplest and con-ventional method in cluster analysis is the k-means clus-tering algorithm+is algorithm segregates the given datasetinto two or more clusters [7] +e accuracy of this methodcompletely depends on the selection of the cluster center Itis mandatory to select the optimum cluster center to get abetter result +e Euclidean distance is the general measureto segregate the dataset [8] Pixels are assigned to an indi-vidual cluster based on the Euclidean distance +e objectivefunction used in this algorithm is
J(v) 1113944C
i11113944
Ci
j1xi minus vj
1113874 11138752 (1)
where xi are the pixels vj are the cluster centers xi minus vj isthe Euclidean distance between xi and vjCi is the number ofdata points for the ith cluster and C is the number of clustercenters [9] +e steps followed for k-means clustering aregiven in Algorithm 4
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 13 Resultant images by k-median clustering
Computational and Mathematical Methods in Medicine 11
25 k-Median Clustering Algorithm +is is also a clusteringalgorithm slightly modified from the k-means algorithm Incentroid calculation instead of calculating the mean valuethe median value is considered +is algorithm significantlyreduces the error since there is no squared operation as inthe calculation of the Euclidean distance +e clustersformed by this method are more compact As an alternatethis approach uses the Lloyd-style iteration +e steps fol-lowed for k-median clustering are given in Algorithm 5 [10]
26 Particle Swarm Optimization PSO is a metaheuristicalgorithm used efficiently in medical image analysis [11] Itmimics the social behavior of the birds searching for food [12]+e fundamental idea of PSO is sharing and communicatingthe information In this approach each particle has initialposition and velocity Based on the fitness value the velocity
and position are updated +e relevant two equations in PSOto update the position and velocity are as follows [11 12]
v(t + 1) v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minus x(t)]
x(t + 1) x(t) + v(t + 1)
(2)
where r1 and r2 are the random numbers and the accel-eration coefficients c1 and c2 are two positive constants+e success of PSO relies on the fitness function +efollowing fitness function has been used for the presentstudy
maximizef 1113944n
i1
intercluster distanceintracluster distance
(3)
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 14 Resultant images by the PSO algorithm
12 Computational and Mathematical Methods in Medicine
where n is the number of clusters +e steps followed for theparticle swarm optimization are shown in Algorithm 6
27 Inertia-Weighted Particle Swarm Optimization +eexploration and exploitation in PSO are based on the inertiaweight +e basic PSO presented by Eberhart and Kennedyin 1995 has no inertia weight In 1998 Shi and Eberhartintroduced the concept of inertia weight by adding constantinertia weight +ey stated that a significant inertia weightfacilitates a global search while a small inertia weight fa-cilitates a local search [14] +is enhances the convergencerate and reduces the number of iterations Inertia weight lessthan 1 in general improves the results +e used methodimproves the convergence rate and saves the time taken andsome iterations
+e resulting velocity update equation becomes
v(t + 1) wlowast v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minusx(t)](4)
where w is the inertia weight with constant inertia weight w
07 and random inertia weight w 05 + rand()2
28 Guaranteed Convergence Particle Swarm Optimization+e GCPSO focuses on a new particle which deals with thecurrent best position in the region In this task this particleis treated as a member of the swarm and the velocity updateequation for this new particle is given as follows [15]
vφ(t + 1) xφ(t) + pbest(t) + ωvφ(t) + ρ(t)(1minus 2r) (5)
+e search ability is increased by the social part +is willimprove the random search in the area around the gbestposition +e random vector and diameter of the search area
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 15 Resultant images by IWPSO algorithm clustering
Computational and Mathematical Methods in Medicine 13
are r and ρ(t) respectively +e range of the random vectorlies between 0 and 1 +e diameter of the search area can beupdated using the following equation
ρ(t + 1)
2ρ(t) successesgt sc
115
1113874 1113875ρ(t) failuresgt fc
ρ(t) otherwise
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎩
(6)
where the terms successes and failures are defined asthe number of consecutive successes and failures re-spectively +e threshold parameters sc and fc are de-termined empirically Since it is hard to obtain a bettervalue in only a few iterations in a high-dimensional searchspace the recommended values are thus sc 15 and fc 5
On some benchmark tests the GCPSO has shown anexcellent performance of locating the minimal of a spaceafter unimodal with only a small amount of particles +esteps to be followed for the GCPSO are shown inAlgorithm 7
3 Performance Measures
Certain performance measures are used to evaluate theresults obtained from medical image segmentation +e listof performance measures used to assess the filter operation isshown in Figure 2 [16] Let If be the image after noise re-duction and I0 be the noisy image
Performance measures used for the evaluation of theresults of the segmentation algorithm are given in Figure 3[17]
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 16 Resultant images by GCPSO algorithm clustering
14 Computational and Mathematical Methods in Medicine
4 Results and Discussion
+e used methods are practically implemented usingMATLAB coding and the results were verified
In the preprocessing stage a comparison was donebetween the performance of median adaptive median andmean filters +e SSI and SMPI values are shown in Table 1and Figures 4 and 5 From the results it is evident that theadaptive median filter has accurate characteristics than themean and median filters for medical image segmentation
+e segmentation accuracy was measured using the truepositive rate true negative rate false positive rate and falsenegative rate by comparing the results from the algorithmwithmanual segmentation results +e practical results of the k-means clustering segmentation algorithm are shown in Table 2
+e practical results of the k-median clustering seg-mentation algorithm are shown in Table 3
+e practical results of the PSO-based segmentationalgorithm are shown in Table 4
+e practical results of the IWPSO segmentation algo-rithm are shown in Table 5
+e practical results of the GCPSO segmentation algo-rithm are shown in Table 6
+e graphical view of the comparison of the true positiverate true negative rate false positive rate and false negativerate for the algorithms used is shown in Figures 6ndash9 It isproved that the true positive and true negative rates are highand false positive and false negative rates are low for theGCPSO algorithm
+e comparative evaluation based on the accuracy of thesegmentation is shown in Table 7 and Figure 10 +e resultsindicate that the GCPSO-based technique has the highestaverage value of accuracy than the other methods
+e resultant images after preprocessing are shown inFigures 11(a) and 11(b)
+e resultant images after segmentation using k-meansclustering are shown in Figure 12
+e resultant images after segmentation using k-medianclustering are shown in Figure 13
+e resultant images after segmentation using the PSOalgorithm are shown in Figure 14
+e resultant images after segmentation using theIWPSO algorithm are shown in Figure 15
+e resultant images after segmentation using theGCPSO algorithm are shown in Figure 16
In an earlier research lung cancer detection was doneusing PSO genetic optimization and SVM algorithm withthe Gabor filter and produced an accuracy of 895 [18]+emethod to detect lung cancer by means of K-NN classifi-cation using the genetic algorithm produced a maximumaccuracy of 90 [19]+e comparative results with respect tothe above-said methods are shown in Table 8
+e graphical comparative analysis between the used andexisting methods is shown in Figure 17
5 Conclusion
In this study various optimization algorithms have beenevaluated to detect the tumor Medical images often need
preprocessing before being subjected to statistical analysis+e adaptive median filter has better results than medianand mean filters because the speckle suppression index andspeckle and mean preservation index values are lower for theadaptive median filter Comparing the five algorithms theaccuracy of the tumor extraction is improved in GCPSOwith the highest accuracy of 958079 and it obtained above90 of precision in all the 20 images It is more accuratewhen compared to the previous method which had an ac-curacy of 90 in 4 out of 10 datasets only In future studiesthe use of more number of optimization algorithms will beincluded to improve the accuracy
Data Availability
+eCTimages data used to support the findings of this studyhave been deposited in the LungCT-Diagnosis repository(doiorg107937K9TCIA2015A6V7JIWX)
Conflicts of Interest
+e authors declare that there are no conflicts of interestregarding the publication of this paper
References
[1] A A Brindha S Indirani and A Srinivasan ldquoLung cancerdetection using SVM algorithm and optimization techniquesrdquoJournal of Chemical and Pharmaceutical Sciences vol 9 no 42016
[2] M Kurkure and A +akare ldquoIntroducing automated systemfor lung cancer detection using Evolutionary ApproachrdquoInternational Journal of Engineering and Computer Sciencevol 5 no 5 pp 16736ndash16739 2016
[3] B Rani A K Goel and R Kaur ldquoA modified approach forlung cancer detection using bacterial forging optimization
868890929496
PSO GA and SVM algorithm
K-NNclassification
using GA
Proposed GCPSOmethod
8950 90
9581
Acc
urac
y
Various algorithms
Comparison between existing and projected methods
Figure 17 Graphical view of accuracy
Table 8 Comparative analysis of accuracy of the projected methodwith various methods
Various methods Accuracy ()PSO GA and SVM algorithm [18] 8950K-NN classification using GA [19] 9000Projected GCPSO method 9581
Computational and Mathematical Methods in Medicine 15
algorithmrdquo International Journal of Scientific Research En-gineering and Technology vol 5 no 1 2016
[4] N Panpaliya N Tadas S Bobade R Aglawe andA Gudadhe ldquoA survey on early detection and prediction oflung cancerrdquo International Journal of Computer Science andMobile Computing vol 4 no 1 pp 175ndash184 2015
[5] G Gupta ldquoAlgorithm for image processing using improvedmedian filter and comparison of mean median and improvedmedian filterrdquo International Journal of Soft Computing andEngineering vol 1 no 5 2011
[6] Tutorial point with digital image processing httpwwwtutorialspointcomdiphistogram_equalizationhtm
[7] K Venkatalakshmi and S M Shalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and K-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Melbourne VIC Australia December2005
[8] K Venkatalakshmi and S M Shalinie ldquoMultispectral imageclassification using modified k-means clusteringrdquo In-ternational Journal on Neural and Mass-Parallel Computingand Information Systems vol 17 no 2 pp 113ndash120 2007
[9] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercy Shalinie ldquoMultispectral image clustering using en-hanced genetic k-means algorithmrdquo Information TechnologyJournal vol 6 no 4 pp 554ndash560 2007
[10] P I Dalatu ldquoTime complexity of k-means and k-mediansclustering algorithms in outliers detectionrdquo Global Journal ofPure and Applied Mathematics vol 12 no 5 pp 4405ndash44182016
[11] K Senthilkumar K Venkatalakshmi K Karthikeyan andP Kathirkamasundari ldquoAn efficient method for segmentingdigital image using a hybrid model of particle swarm opti-mization and artificial bee colony algorithmrdquo InternationalJournal of Applied Engineering Research vol 10 pp 444ndash4492015
[12] K Venkatalakshmi and S Mercyshalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and k-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Bangalore India December 2005
[13] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercyshalinie ldquoA customized particle swarm optimizationfor classification of multispectral imagery based on featurefusionrdquo International Arab Journal of Information Technologyvol 5 no 4 pp 327ndash333 2008
[14] J C Bansal and P K Singh ldquoInertia weight strategies inparticle swarm optimizationrdquo in Proceedings of IEEE WorldCongress on Nature and Biologically Inspired Computing pp640ndash647 Salamanca Spain October 2011
[15] P K Patel V Sharma and K Gupta ldquoGuaranteed conver-gence particle swarm optimization using Personal Bestrdquo In-ternational Journal of Computer Applications vol 73 no 7pp 6ndash10 2013
[16] X Wang L Ge and X Li ldquoEvaluation of filters for Envi-satAsar speckle suppression in pasture areardquo ISPRS Annals ofthe Photogrammetry Remote Sensing and Spatial InformationSciences vol 1ndash7 pp 341ndash346 2012
[17] H Nagaveena D Devaraj and S C Prasanna KumarldquoVessels segmentation in diabetic retinopathy by adaptivemedian thresholdingrdquo International Journal of Science andTechnology vol 1 no 1 pp 17ndash22 2013
[18] A Asuntha N Singh and A Srinivasan ldquoPSO genetic op-timization and SVM algorithm used for lung cancer
detectionrdquo Journal of Chemical and Pharmaceutical Researchvol 8 no 6 pp 351ndash359 2016
[19] P Bhuvaneswari and A Brintha+erese ldquoDetection of cancerin lung with K-NN classification using genetic algorithmrdquoInternational Conference on Nanomaterials and Technologiesvol 10 pp 433ndash440 2014
16 Computational and Mathematical Methods in Medicine
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom
Table 4 Statistical results from the PSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 875413 901196 98804 124587 890624Image 2 859612 859612 85479 85479 889689Image 3 827919 898314 101686 172081 864850Image 4 811271 887838 112162 188729 849967Image 5 826343 873995 126005 173657 850299Image 6 818996 852900 147100 181004 835581Image 7 817281 800949 199051 182719 809438Image 8 804182 800721 199279 195818 802571Image 9 822573 811450 188550 177427 817340Image 10 808521 803433 196567 191479 806182Image 11 822198 809837 190163 177802 816421Image 12 846322 815070 184930 153678 831347Image 13 826283 811153 188847 173617 819351Image 14 809338 809090 190910 190662 809226Image 15 818790 811729 188271 181210 815586Image 16 808120 814222 185778 191880 810836Image 17 808582 818136 181864 191418 812824Image 18 791387 850084 149916 208613 818114Image 19 794655 851570 148430 205345 820954Image 20 777826 853446 146554 222174 811744
Table 5 Statistical results from the IWPSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 874649 903272 96728 125351 889813Image 2 861950 898126 101874 138050 881810Image 3 829347 889622 110378 170653 861018Image 4 814285 872013 127987 185715 843584Image 5 829023 863940 136060 170977 846631Image 6 821065 845145 154855 178935 832876Image 7 821361 791744 208256 178639 807064Image 8 803274 803855 196145 196726 803544Image 9 824185 803924 196076 175815 814631Image 10 810769 795965 204035 189231 803943Image 11 822198 797401 202599 174299 812411Image 12 846322 811390 188610 152172 830334Image 13 826283 807231 173596 173596 818905Image 14 809338 808231 191769 190328 809025Image 15 818790 815899 184101 182694 816669Image 16 808120 814173 185827 191734 810896Image 17 808582 810677 189323 190166 810209Image 18 791387 849622 150378 208310 818081Image 19 794655 848219 151781 203936 820213Image 20 777684 854281 145719 222316 812033
Table 3 Continued
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 15 813002 800536 199464 186998 807248Image 16 801942 800503 199497 198058 801291Image 17 802984 803756 196244 197016 803332Image 18 786327 855226 144774 213673 817792Image 19 789322 852163 147837 210678 818439Image 20 772752 858000 142000 227248 810973
6 Computational and Mathematical Methods in Medicine
Table 6 Statistical results from the GCPSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 916158 999999 00001 83842 958079Image 2 909563 999999 00001 90437 954782Image 3 888404 999999 00001 111592 944204Image 4 872946 999999 00001 127054 936473Image 5 873583 999999 00001 126417 936792Image 6 861567 999999 00001 138433 930784Image 7 834867 999999 00001 165133 917434Image 8 831082 999999 00001 168918 915541Image 9 842907 999999 00001 157093 921453Image 10 831917 999999 00001 168083 915958Image 11 842122 999999 00001 157878 921061Image 12 857977 999999 00001 142023 928988Image 13 846397 999999 00001 153603 923198Image 14 837442 999999 00001 162558 918721Image 15 843299 999999 00001 156701 921649Image 16 838867 999999 00001 161133 919433Image 17 839061 999999 00001 160939 919531Image 18 846836 999999 00001 153164 923418Image 19 849324 999999 00001 150676 924662Image 20 839867 999999 00001 160124 919938
6500
7000
7500
8000
8500
9000
9500
True
pos
itive
rate
k-meansk-mediansPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
IWPSOGCPSO
Figure 6 Comparative results of the true positive rate value
75
80
85
90
95
100
True
neg
ativ
e rat
e
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 7 Comparative results of the true negative rate value
Computational and Mathematical Methods in Medicine 7
72
77
82
87
92
97
Acc
urac
y
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 10 Comparative results of accuracy
0
5
10
15
20
25
False
pos
itive
rate
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 8 Comparative results of the false positive rate value
0
5
10
15
20
25
False
neg
ativ
e rat
e
k-meansk-mediansPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
IWPSOGCPSO
Figure 9 Comparative results of the false negative rate value
8 Computational and Mathematical Methods in Medicine
Table 7 Statistical comparative result of accuracy
Images k-Means k-Median PSO IWPSO GCPSOImage 1 885937 893672 890624 889813 958079Image 2 882682 885622 889689 881810 954782Image 3 857501 862969 864850 861018 944204Image 4 842186 842695 849967 843584 936473Image 5 839216 845294 850299 846631 936792Image 6 825795 833654 835581 832876 930784Image 7 801519 807832 809438 807064 917434Image 8 790656 798186 802571 803544 915541Image 9 801606 809977 817340 814631 921453Image 10 790378 799611 806182 803943 915958Image 11 799755 806621 816421 812411 921061Image 12 815238 824514 831347 830334 928988Image 13 802806 812545 819351 818905 923198Image 14 793023 799876 809226 809025 918721Image 15 797600 807248 815586 816669 921649Image 16 792121 801291 810836 810896 919433Image 17 792229 803332 812824 810209 919531Image 18 807020 817792 818114 818081 923418Image 19 810816 818439 820954 820213 924662Image 20 801831 810973 811744 812033 919938
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 1
Image 2
Image 3
Image 4
Image 5
Image 6
Image 7
Image 8
Image 9
Image 10
(a)
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
(b)
Figure 11 Resultant images after preprocessing
Computational and Mathematical Methods in Medicine 9
eradicate the incidence of noise content and to improve theimage quality before an examination [4]+is part of work isknown as preprocessing In the preprocessing stage noiseremoval and contrast enhancement are two primary steps Inthe present study the performance results of medianadaptive median and average filters to isolate the presence ofspeckle noise have been compared +e coding for the samehas been implemented using MATLAB Furthermore theimage quality and visual appearance are improved byadaptive histogram equalization+e second stage of work issegmentation +is stage consists of applying five methodsnamely k-means k-median particle swarm optimization(PSO) inertia-weighted particle swarm optimization(IWPSO) and GCPSO +e tumor portion was extractedfrom the segmented results of the above-said five methodsand compared with manual extraction+e results show thatthe GCPSO-based segmentation has more accuracy than theothers Figure 1 depicts the process of operation for thepresent study
21 Median and Adaptive Median Filters +e median filterremoves the noise and retains the sharpness of the imageAccordance to the name each pixel is replaced by themedian value from the neighborhood pixels A 3 times 3 windowis used in this filter [5] +is is one of the best filters amongconventional filters which remove the speckle noise +esteps followed to construct the median filter are given inAlgorithm 1
Spatial processing to preserve the edge detail and toeliminate nonimpulsive noise by the adaptive median filterplays a vital role +e small structure in the image and edgesare retained by the adaptive median filter In the adaptivemedian filter the window size varies with respect to eachpixel
22 Average Filter +is is a simple filter which removesthe spatial noise from a digital image +e presence ofspatial noise is mainly due to the data acquisition process
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 12 Resultant images by k-means clustering
10 Computational and Mathematical Methods in Medicine
+e neighborhood mean value is measured for each andevery pixel and is replaced by the corresponding meanvalue +is process is repeated for every pixel in the image[5] All the pixels in the digital image are modified bysliding the operator over the entire range of pixels +esteps followed for the average filter are given inAlgorithm 2
23 Histogram Equalization Image enhancement is thetechnique which is used to improve the image quality Forbetter understanding and analysis it is mandatory to en-hance the contrast of medical images +e conventionalmethod used for this operation is histogram equalization Aminor adjustment on the intensity of image pixels is donein this method Each pixel is mapped to intensity pro-portional to its rank in the surrounding pixels +e stepsfollowed for histogram equalization are given in Algo-rithm 3 [6]
24 k-Means Clustering Algorithm +e simplest and con-ventional method in cluster analysis is the k-means clus-tering algorithm+is algorithm segregates the given datasetinto two or more clusters [7] +e accuracy of this methodcompletely depends on the selection of the cluster center Itis mandatory to select the optimum cluster center to get abetter result +e Euclidean distance is the general measureto segregate the dataset [8] Pixels are assigned to an indi-vidual cluster based on the Euclidean distance +e objectivefunction used in this algorithm is
J(v) 1113944C
i11113944
Ci
j1xi minus vj
1113874 11138752 (1)
where xi are the pixels vj are the cluster centers xi minus vj isthe Euclidean distance between xi and vjCi is the number ofdata points for the ith cluster and C is the number of clustercenters [9] +e steps followed for k-means clustering aregiven in Algorithm 4
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 13 Resultant images by k-median clustering
Computational and Mathematical Methods in Medicine 11
25 k-Median Clustering Algorithm +is is also a clusteringalgorithm slightly modified from the k-means algorithm Incentroid calculation instead of calculating the mean valuethe median value is considered +is algorithm significantlyreduces the error since there is no squared operation as inthe calculation of the Euclidean distance +e clustersformed by this method are more compact As an alternatethis approach uses the Lloyd-style iteration +e steps fol-lowed for k-median clustering are given in Algorithm 5 [10]
26 Particle Swarm Optimization PSO is a metaheuristicalgorithm used efficiently in medical image analysis [11] Itmimics the social behavior of the birds searching for food [12]+e fundamental idea of PSO is sharing and communicatingthe information In this approach each particle has initialposition and velocity Based on the fitness value the velocity
and position are updated +e relevant two equations in PSOto update the position and velocity are as follows [11 12]
v(t + 1) v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minus x(t)]
x(t + 1) x(t) + v(t + 1)
(2)
where r1 and r2 are the random numbers and the accel-eration coefficients c1 and c2 are two positive constants+e success of PSO relies on the fitness function +efollowing fitness function has been used for the presentstudy
maximizef 1113944n
i1
intercluster distanceintracluster distance
(3)
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 14 Resultant images by the PSO algorithm
12 Computational and Mathematical Methods in Medicine
where n is the number of clusters +e steps followed for theparticle swarm optimization are shown in Algorithm 6
27 Inertia-Weighted Particle Swarm Optimization +eexploration and exploitation in PSO are based on the inertiaweight +e basic PSO presented by Eberhart and Kennedyin 1995 has no inertia weight In 1998 Shi and Eberhartintroduced the concept of inertia weight by adding constantinertia weight +ey stated that a significant inertia weightfacilitates a global search while a small inertia weight fa-cilitates a local search [14] +is enhances the convergencerate and reduces the number of iterations Inertia weight lessthan 1 in general improves the results +e used methodimproves the convergence rate and saves the time taken andsome iterations
+e resulting velocity update equation becomes
v(t + 1) wlowast v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minusx(t)](4)
where w is the inertia weight with constant inertia weight w
07 and random inertia weight w 05 + rand()2
28 Guaranteed Convergence Particle Swarm Optimization+e GCPSO focuses on a new particle which deals with thecurrent best position in the region In this task this particleis treated as a member of the swarm and the velocity updateequation for this new particle is given as follows [15]
vφ(t + 1) xφ(t) + pbest(t) + ωvφ(t) + ρ(t)(1minus 2r) (5)
+e search ability is increased by the social part +is willimprove the random search in the area around the gbestposition +e random vector and diameter of the search area
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 15 Resultant images by IWPSO algorithm clustering
Computational and Mathematical Methods in Medicine 13
are r and ρ(t) respectively +e range of the random vectorlies between 0 and 1 +e diameter of the search area can beupdated using the following equation
ρ(t + 1)
2ρ(t) successesgt sc
115
1113874 1113875ρ(t) failuresgt fc
ρ(t) otherwise
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎩
(6)
where the terms successes and failures are defined asthe number of consecutive successes and failures re-spectively +e threshold parameters sc and fc are de-termined empirically Since it is hard to obtain a bettervalue in only a few iterations in a high-dimensional searchspace the recommended values are thus sc 15 and fc 5
On some benchmark tests the GCPSO has shown anexcellent performance of locating the minimal of a spaceafter unimodal with only a small amount of particles +esteps to be followed for the GCPSO are shown inAlgorithm 7
3 Performance Measures
Certain performance measures are used to evaluate theresults obtained from medical image segmentation +e listof performance measures used to assess the filter operation isshown in Figure 2 [16] Let If be the image after noise re-duction and I0 be the noisy image
Performance measures used for the evaluation of theresults of the segmentation algorithm are given in Figure 3[17]
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 16 Resultant images by GCPSO algorithm clustering
14 Computational and Mathematical Methods in Medicine
4 Results and Discussion
+e used methods are practically implemented usingMATLAB coding and the results were verified
In the preprocessing stage a comparison was donebetween the performance of median adaptive median andmean filters +e SSI and SMPI values are shown in Table 1and Figures 4 and 5 From the results it is evident that theadaptive median filter has accurate characteristics than themean and median filters for medical image segmentation
+e segmentation accuracy was measured using the truepositive rate true negative rate false positive rate and falsenegative rate by comparing the results from the algorithmwithmanual segmentation results +e practical results of the k-means clustering segmentation algorithm are shown in Table 2
+e practical results of the k-median clustering seg-mentation algorithm are shown in Table 3
+e practical results of the PSO-based segmentationalgorithm are shown in Table 4
+e practical results of the IWPSO segmentation algo-rithm are shown in Table 5
+e practical results of the GCPSO segmentation algo-rithm are shown in Table 6
+e graphical view of the comparison of the true positiverate true negative rate false positive rate and false negativerate for the algorithms used is shown in Figures 6ndash9 It isproved that the true positive and true negative rates are highand false positive and false negative rates are low for theGCPSO algorithm
+e comparative evaluation based on the accuracy of thesegmentation is shown in Table 7 and Figure 10 +e resultsindicate that the GCPSO-based technique has the highestaverage value of accuracy than the other methods
+e resultant images after preprocessing are shown inFigures 11(a) and 11(b)
+e resultant images after segmentation using k-meansclustering are shown in Figure 12
+e resultant images after segmentation using k-medianclustering are shown in Figure 13
+e resultant images after segmentation using the PSOalgorithm are shown in Figure 14
+e resultant images after segmentation using theIWPSO algorithm are shown in Figure 15
+e resultant images after segmentation using theGCPSO algorithm are shown in Figure 16
In an earlier research lung cancer detection was doneusing PSO genetic optimization and SVM algorithm withthe Gabor filter and produced an accuracy of 895 [18]+emethod to detect lung cancer by means of K-NN classifi-cation using the genetic algorithm produced a maximumaccuracy of 90 [19]+e comparative results with respect tothe above-said methods are shown in Table 8
+e graphical comparative analysis between the used andexisting methods is shown in Figure 17
5 Conclusion
In this study various optimization algorithms have beenevaluated to detect the tumor Medical images often need
preprocessing before being subjected to statistical analysis+e adaptive median filter has better results than medianand mean filters because the speckle suppression index andspeckle and mean preservation index values are lower for theadaptive median filter Comparing the five algorithms theaccuracy of the tumor extraction is improved in GCPSOwith the highest accuracy of 958079 and it obtained above90 of precision in all the 20 images It is more accuratewhen compared to the previous method which had an ac-curacy of 90 in 4 out of 10 datasets only In future studiesthe use of more number of optimization algorithms will beincluded to improve the accuracy
Data Availability
+eCTimages data used to support the findings of this studyhave been deposited in the LungCT-Diagnosis repository(doiorg107937K9TCIA2015A6V7JIWX)
Conflicts of Interest
+e authors declare that there are no conflicts of interestregarding the publication of this paper
References
[1] A A Brindha S Indirani and A Srinivasan ldquoLung cancerdetection using SVM algorithm and optimization techniquesrdquoJournal of Chemical and Pharmaceutical Sciences vol 9 no 42016
[2] M Kurkure and A +akare ldquoIntroducing automated systemfor lung cancer detection using Evolutionary ApproachrdquoInternational Journal of Engineering and Computer Sciencevol 5 no 5 pp 16736ndash16739 2016
[3] B Rani A K Goel and R Kaur ldquoA modified approach forlung cancer detection using bacterial forging optimization
868890929496
PSO GA and SVM algorithm
K-NNclassification
using GA
Proposed GCPSOmethod
8950 90
9581
Acc
urac
y
Various algorithms
Comparison between existing and projected methods
Figure 17 Graphical view of accuracy
Table 8 Comparative analysis of accuracy of the projected methodwith various methods
Various methods Accuracy ()PSO GA and SVM algorithm [18] 8950K-NN classification using GA [19] 9000Projected GCPSO method 9581
Computational and Mathematical Methods in Medicine 15
algorithmrdquo International Journal of Scientific Research En-gineering and Technology vol 5 no 1 2016
[4] N Panpaliya N Tadas S Bobade R Aglawe andA Gudadhe ldquoA survey on early detection and prediction oflung cancerrdquo International Journal of Computer Science andMobile Computing vol 4 no 1 pp 175ndash184 2015
[5] G Gupta ldquoAlgorithm for image processing using improvedmedian filter and comparison of mean median and improvedmedian filterrdquo International Journal of Soft Computing andEngineering vol 1 no 5 2011
[6] Tutorial point with digital image processing httpwwwtutorialspointcomdiphistogram_equalizationhtm
[7] K Venkatalakshmi and S M Shalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and K-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Melbourne VIC Australia December2005
[8] K Venkatalakshmi and S M Shalinie ldquoMultispectral imageclassification using modified k-means clusteringrdquo In-ternational Journal on Neural and Mass-Parallel Computingand Information Systems vol 17 no 2 pp 113ndash120 2007
[9] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercy Shalinie ldquoMultispectral image clustering using en-hanced genetic k-means algorithmrdquo Information TechnologyJournal vol 6 no 4 pp 554ndash560 2007
[10] P I Dalatu ldquoTime complexity of k-means and k-mediansclustering algorithms in outliers detectionrdquo Global Journal ofPure and Applied Mathematics vol 12 no 5 pp 4405ndash44182016
[11] K Senthilkumar K Venkatalakshmi K Karthikeyan andP Kathirkamasundari ldquoAn efficient method for segmentingdigital image using a hybrid model of particle swarm opti-mization and artificial bee colony algorithmrdquo InternationalJournal of Applied Engineering Research vol 10 pp 444ndash4492015
[12] K Venkatalakshmi and S Mercyshalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and k-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Bangalore India December 2005
[13] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercyshalinie ldquoA customized particle swarm optimizationfor classification of multispectral imagery based on featurefusionrdquo International Arab Journal of Information Technologyvol 5 no 4 pp 327ndash333 2008
[14] J C Bansal and P K Singh ldquoInertia weight strategies inparticle swarm optimizationrdquo in Proceedings of IEEE WorldCongress on Nature and Biologically Inspired Computing pp640ndash647 Salamanca Spain October 2011
[15] P K Patel V Sharma and K Gupta ldquoGuaranteed conver-gence particle swarm optimization using Personal Bestrdquo In-ternational Journal of Computer Applications vol 73 no 7pp 6ndash10 2013
[16] X Wang L Ge and X Li ldquoEvaluation of filters for Envi-satAsar speckle suppression in pasture areardquo ISPRS Annals ofthe Photogrammetry Remote Sensing and Spatial InformationSciences vol 1ndash7 pp 341ndash346 2012
[17] H Nagaveena D Devaraj and S C Prasanna KumarldquoVessels segmentation in diabetic retinopathy by adaptivemedian thresholdingrdquo International Journal of Science andTechnology vol 1 no 1 pp 17ndash22 2013
[18] A Asuntha N Singh and A Srinivasan ldquoPSO genetic op-timization and SVM algorithm used for lung cancer
detectionrdquo Journal of Chemical and Pharmaceutical Researchvol 8 no 6 pp 351ndash359 2016
[19] P Bhuvaneswari and A Brintha+erese ldquoDetection of cancerin lung with K-NN classification using genetic algorithmrdquoInternational Conference on Nanomaterials and Technologiesvol 10 pp 433ndash440 2014
16 Computational and Mathematical Methods in Medicine
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom
Table 6 Statistical results from the GCPSO algorithm
Images True positive rate True negative rate False positive rate False negative rate AccuracyImage 1 916158 999999 00001 83842 958079Image 2 909563 999999 00001 90437 954782Image 3 888404 999999 00001 111592 944204Image 4 872946 999999 00001 127054 936473Image 5 873583 999999 00001 126417 936792Image 6 861567 999999 00001 138433 930784Image 7 834867 999999 00001 165133 917434Image 8 831082 999999 00001 168918 915541Image 9 842907 999999 00001 157093 921453Image 10 831917 999999 00001 168083 915958Image 11 842122 999999 00001 157878 921061Image 12 857977 999999 00001 142023 928988Image 13 846397 999999 00001 153603 923198Image 14 837442 999999 00001 162558 918721Image 15 843299 999999 00001 156701 921649Image 16 838867 999999 00001 161133 919433Image 17 839061 999999 00001 160939 919531Image 18 846836 999999 00001 153164 923418Image 19 849324 999999 00001 150676 924662Image 20 839867 999999 00001 160124 919938
6500
7000
7500
8000
8500
9000
9500
True
pos
itive
rate
k-meansk-mediansPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
IWPSOGCPSO
Figure 6 Comparative results of the true positive rate value
75
80
85
90
95
100
True
neg
ativ
e rat
e
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 7 Comparative results of the true negative rate value
Computational and Mathematical Methods in Medicine 7
72
77
82
87
92
97
Acc
urac
y
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 10 Comparative results of accuracy
0
5
10
15
20
25
False
pos
itive
rate
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 8 Comparative results of the false positive rate value
0
5
10
15
20
25
False
neg
ativ
e rat
e
k-meansk-mediansPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
IWPSOGCPSO
Figure 9 Comparative results of the false negative rate value
8 Computational and Mathematical Methods in Medicine
Table 7 Statistical comparative result of accuracy
Images k-Means k-Median PSO IWPSO GCPSOImage 1 885937 893672 890624 889813 958079Image 2 882682 885622 889689 881810 954782Image 3 857501 862969 864850 861018 944204Image 4 842186 842695 849967 843584 936473Image 5 839216 845294 850299 846631 936792Image 6 825795 833654 835581 832876 930784Image 7 801519 807832 809438 807064 917434Image 8 790656 798186 802571 803544 915541Image 9 801606 809977 817340 814631 921453Image 10 790378 799611 806182 803943 915958Image 11 799755 806621 816421 812411 921061Image 12 815238 824514 831347 830334 928988Image 13 802806 812545 819351 818905 923198Image 14 793023 799876 809226 809025 918721Image 15 797600 807248 815586 816669 921649Image 16 792121 801291 810836 810896 919433Image 17 792229 803332 812824 810209 919531Image 18 807020 817792 818114 818081 923418Image 19 810816 818439 820954 820213 924662Image 20 801831 810973 811744 812033 919938
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 1
Image 2
Image 3
Image 4
Image 5
Image 6
Image 7
Image 8
Image 9
Image 10
(a)
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
(b)
Figure 11 Resultant images after preprocessing
Computational and Mathematical Methods in Medicine 9
eradicate the incidence of noise content and to improve theimage quality before an examination [4]+is part of work isknown as preprocessing In the preprocessing stage noiseremoval and contrast enhancement are two primary steps Inthe present study the performance results of medianadaptive median and average filters to isolate the presence ofspeckle noise have been compared +e coding for the samehas been implemented using MATLAB Furthermore theimage quality and visual appearance are improved byadaptive histogram equalization+e second stage of work issegmentation +is stage consists of applying five methodsnamely k-means k-median particle swarm optimization(PSO) inertia-weighted particle swarm optimization(IWPSO) and GCPSO +e tumor portion was extractedfrom the segmented results of the above-said five methodsand compared with manual extraction+e results show thatthe GCPSO-based segmentation has more accuracy than theothers Figure 1 depicts the process of operation for thepresent study
21 Median and Adaptive Median Filters +e median filterremoves the noise and retains the sharpness of the imageAccordance to the name each pixel is replaced by themedian value from the neighborhood pixels A 3 times 3 windowis used in this filter [5] +is is one of the best filters amongconventional filters which remove the speckle noise +esteps followed to construct the median filter are given inAlgorithm 1
Spatial processing to preserve the edge detail and toeliminate nonimpulsive noise by the adaptive median filterplays a vital role +e small structure in the image and edgesare retained by the adaptive median filter In the adaptivemedian filter the window size varies with respect to eachpixel
22 Average Filter +is is a simple filter which removesthe spatial noise from a digital image +e presence ofspatial noise is mainly due to the data acquisition process
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 12 Resultant images by k-means clustering
10 Computational and Mathematical Methods in Medicine
+e neighborhood mean value is measured for each andevery pixel and is replaced by the corresponding meanvalue +is process is repeated for every pixel in the image[5] All the pixels in the digital image are modified bysliding the operator over the entire range of pixels +esteps followed for the average filter are given inAlgorithm 2
23 Histogram Equalization Image enhancement is thetechnique which is used to improve the image quality Forbetter understanding and analysis it is mandatory to en-hance the contrast of medical images +e conventionalmethod used for this operation is histogram equalization Aminor adjustment on the intensity of image pixels is donein this method Each pixel is mapped to intensity pro-portional to its rank in the surrounding pixels +e stepsfollowed for histogram equalization are given in Algo-rithm 3 [6]
24 k-Means Clustering Algorithm +e simplest and con-ventional method in cluster analysis is the k-means clus-tering algorithm+is algorithm segregates the given datasetinto two or more clusters [7] +e accuracy of this methodcompletely depends on the selection of the cluster center Itis mandatory to select the optimum cluster center to get abetter result +e Euclidean distance is the general measureto segregate the dataset [8] Pixels are assigned to an indi-vidual cluster based on the Euclidean distance +e objectivefunction used in this algorithm is
J(v) 1113944C
i11113944
Ci
j1xi minus vj
1113874 11138752 (1)
where xi are the pixels vj are the cluster centers xi minus vj isthe Euclidean distance between xi and vjCi is the number ofdata points for the ith cluster and C is the number of clustercenters [9] +e steps followed for k-means clustering aregiven in Algorithm 4
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 13 Resultant images by k-median clustering
Computational and Mathematical Methods in Medicine 11
25 k-Median Clustering Algorithm +is is also a clusteringalgorithm slightly modified from the k-means algorithm Incentroid calculation instead of calculating the mean valuethe median value is considered +is algorithm significantlyreduces the error since there is no squared operation as inthe calculation of the Euclidean distance +e clustersformed by this method are more compact As an alternatethis approach uses the Lloyd-style iteration +e steps fol-lowed for k-median clustering are given in Algorithm 5 [10]
26 Particle Swarm Optimization PSO is a metaheuristicalgorithm used efficiently in medical image analysis [11] Itmimics the social behavior of the birds searching for food [12]+e fundamental idea of PSO is sharing and communicatingthe information In this approach each particle has initialposition and velocity Based on the fitness value the velocity
and position are updated +e relevant two equations in PSOto update the position and velocity are as follows [11 12]
v(t + 1) v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minus x(t)]
x(t + 1) x(t) + v(t + 1)
(2)
where r1 and r2 are the random numbers and the accel-eration coefficients c1 and c2 are two positive constants+e success of PSO relies on the fitness function +efollowing fitness function has been used for the presentstudy
maximizef 1113944n
i1
intercluster distanceintracluster distance
(3)
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 14 Resultant images by the PSO algorithm
12 Computational and Mathematical Methods in Medicine
where n is the number of clusters +e steps followed for theparticle swarm optimization are shown in Algorithm 6
27 Inertia-Weighted Particle Swarm Optimization +eexploration and exploitation in PSO are based on the inertiaweight +e basic PSO presented by Eberhart and Kennedyin 1995 has no inertia weight In 1998 Shi and Eberhartintroduced the concept of inertia weight by adding constantinertia weight +ey stated that a significant inertia weightfacilitates a global search while a small inertia weight fa-cilitates a local search [14] +is enhances the convergencerate and reduces the number of iterations Inertia weight lessthan 1 in general improves the results +e used methodimproves the convergence rate and saves the time taken andsome iterations
+e resulting velocity update equation becomes
v(t + 1) wlowast v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minusx(t)](4)
where w is the inertia weight with constant inertia weight w
07 and random inertia weight w 05 + rand()2
28 Guaranteed Convergence Particle Swarm Optimization+e GCPSO focuses on a new particle which deals with thecurrent best position in the region In this task this particleis treated as a member of the swarm and the velocity updateequation for this new particle is given as follows [15]
vφ(t + 1) xφ(t) + pbest(t) + ωvφ(t) + ρ(t)(1minus 2r) (5)
+e search ability is increased by the social part +is willimprove the random search in the area around the gbestposition +e random vector and diameter of the search area
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 15 Resultant images by IWPSO algorithm clustering
Computational and Mathematical Methods in Medicine 13
are r and ρ(t) respectively +e range of the random vectorlies between 0 and 1 +e diameter of the search area can beupdated using the following equation
ρ(t + 1)
2ρ(t) successesgt sc
115
1113874 1113875ρ(t) failuresgt fc
ρ(t) otherwise
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎩
(6)
where the terms successes and failures are defined asthe number of consecutive successes and failures re-spectively +e threshold parameters sc and fc are de-termined empirically Since it is hard to obtain a bettervalue in only a few iterations in a high-dimensional searchspace the recommended values are thus sc 15 and fc 5
On some benchmark tests the GCPSO has shown anexcellent performance of locating the minimal of a spaceafter unimodal with only a small amount of particles +esteps to be followed for the GCPSO are shown inAlgorithm 7
3 Performance Measures
Certain performance measures are used to evaluate theresults obtained from medical image segmentation +e listof performance measures used to assess the filter operation isshown in Figure 2 [16] Let If be the image after noise re-duction and I0 be the noisy image
Performance measures used for the evaluation of theresults of the segmentation algorithm are given in Figure 3[17]
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 16 Resultant images by GCPSO algorithm clustering
14 Computational and Mathematical Methods in Medicine
4 Results and Discussion
+e used methods are practically implemented usingMATLAB coding and the results were verified
In the preprocessing stage a comparison was donebetween the performance of median adaptive median andmean filters +e SSI and SMPI values are shown in Table 1and Figures 4 and 5 From the results it is evident that theadaptive median filter has accurate characteristics than themean and median filters for medical image segmentation
+e segmentation accuracy was measured using the truepositive rate true negative rate false positive rate and falsenegative rate by comparing the results from the algorithmwithmanual segmentation results +e practical results of the k-means clustering segmentation algorithm are shown in Table 2
+e practical results of the k-median clustering seg-mentation algorithm are shown in Table 3
+e practical results of the PSO-based segmentationalgorithm are shown in Table 4
+e practical results of the IWPSO segmentation algo-rithm are shown in Table 5
+e practical results of the GCPSO segmentation algo-rithm are shown in Table 6
+e graphical view of the comparison of the true positiverate true negative rate false positive rate and false negativerate for the algorithms used is shown in Figures 6ndash9 It isproved that the true positive and true negative rates are highand false positive and false negative rates are low for theGCPSO algorithm
+e comparative evaluation based on the accuracy of thesegmentation is shown in Table 7 and Figure 10 +e resultsindicate that the GCPSO-based technique has the highestaverage value of accuracy than the other methods
+e resultant images after preprocessing are shown inFigures 11(a) and 11(b)
+e resultant images after segmentation using k-meansclustering are shown in Figure 12
+e resultant images after segmentation using k-medianclustering are shown in Figure 13
+e resultant images after segmentation using the PSOalgorithm are shown in Figure 14
+e resultant images after segmentation using theIWPSO algorithm are shown in Figure 15
+e resultant images after segmentation using theGCPSO algorithm are shown in Figure 16
In an earlier research lung cancer detection was doneusing PSO genetic optimization and SVM algorithm withthe Gabor filter and produced an accuracy of 895 [18]+emethod to detect lung cancer by means of K-NN classifi-cation using the genetic algorithm produced a maximumaccuracy of 90 [19]+e comparative results with respect tothe above-said methods are shown in Table 8
+e graphical comparative analysis between the used andexisting methods is shown in Figure 17
5 Conclusion
In this study various optimization algorithms have beenevaluated to detect the tumor Medical images often need
preprocessing before being subjected to statistical analysis+e adaptive median filter has better results than medianand mean filters because the speckle suppression index andspeckle and mean preservation index values are lower for theadaptive median filter Comparing the five algorithms theaccuracy of the tumor extraction is improved in GCPSOwith the highest accuracy of 958079 and it obtained above90 of precision in all the 20 images It is more accuratewhen compared to the previous method which had an ac-curacy of 90 in 4 out of 10 datasets only In future studiesthe use of more number of optimization algorithms will beincluded to improve the accuracy
Data Availability
+eCTimages data used to support the findings of this studyhave been deposited in the LungCT-Diagnosis repository(doiorg107937K9TCIA2015A6V7JIWX)
Conflicts of Interest
+e authors declare that there are no conflicts of interestregarding the publication of this paper
References
[1] A A Brindha S Indirani and A Srinivasan ldquoLung cancerdetection using SVM algorithm and optimization techniquesrdquoJournal of Chemical and Pharmaceutical Sciences vol 9 no 42016
[2] M Kurkure and A +akare ldquoIntroducing automated systemfor lung cancer detection using Evolutionary ApproachrdquoInternational Journal of Engineering and Computer Sciencevol 5 no 5 pp 16736ndash16739 2016
[3] B Rani A K Goel and R Kaur ldquoA modified approach forlung cancer detection using bacterial forging optimization
868890929496
PSO GA and SVM algorithm
K-NNclassification
using GA
Proposed GCPSOmethod
8950 90
9581
Acc
urac
y
Various algorithms
Comparison between existing and projected methods
Figure 17 Graphical view of accuracy
Table 8 Comparative analysis of accuracy of the projected methodwith various methods
Various methods Accuracy ()PSO GA and SVM algorithm [18] 8950K-NN classification using GA [19] 9000Projected GCPSO method 9581
Computational and Mathematical Methods in Medicine 15
algorithmrdquo International Journal of Scientific Research En-gineering and Technology vol 5 no 1 2016
[4] N Panpaliya N Tadas S Bobade R Aglawe andA Gudadhe ldquoA survey on early detection and prediction oflung cancerrdquo International Journal of Computer Science andMobile Computing vol 4 no 1 pp 175ndash184 2015
[5] G Gupta ldquoAlgorithm for image processing using improvedmedian filter and comparison of mean median and improvedmedian filterrdquo International Journal of Soft Computing andEngineering vol 1 no 5 2011
[6] Tutorial point with digital image processing httpwwwtutorialspointcomdiphistogram_equalizationhtm
[7] K Venkatalakshmi and S M Shalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and K-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Melbourne VIC Australia December2005
[8] K Venkatalakshmi and S M Shalinie ldquoMultispectral imageclassification using modified k-means clusteringrdquo In-ternational Journal on Neural and Mass-Parallel Computingand Information Systems vol 17 no 2 pp 113ndash120 2007
[9] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercy Shalinie ldquoMultispectral image clustering using en-hanced genetic k-means algorithmrdquo Information TechnologyJournal vol 6 no 4 pp 554ndash560 2007
[10] P I Dalatu ldquoTime complexity of k-means and k-mediansclustering algorithms in outliers detectionrdquo Global Journal ofPure and Applied Mathematics vol 12 no 5 pp 4405ndash44182016
[11] K Senthilkumar K Venkatalakshmi K Karthikeyan andP Kathirkamasundari ldquoAn efficient method for segmentingdigital image using a hybrid model of particle swarm opti-mization and artificial bee colony algorithmrdquo InternationalJournal of Applied Engineering Research vol 10 pp 444ndash4492015
[12] K Venkatalakshmi and S Mercyshalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and k-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Bangalore India December 2005
[13] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercyshalinie ldquoA customized particle swarm optimizationfor classification of multispectral imagery based on featurefusionrdquo International Arab Journal of Information Technologyvol 5 no 4 pp 327ndash333 2008
[14] J C Bansal and P K Singh ldquoInertia weight strategies inparticle swarm optimizationrdquo in Proceedings of IEEE WorldCongress on Nature and Biologically Inspired Computing pp640ndash647 Salamanca Spain October 2011
[15] P K Patel V Sharma and K Gupta ldquoGuaranteed conver-gence particle swarm optimization using Personal Bestrdquo In-ternational Journal of Computer Applications vol 73 no 7pp 6ndash10 2013
[16] X Wang L Ge and X Li ldquoEvaluation of filters for Envi-satAsar speckle suppression in pasture areardquo ISPRS Annals ofthe Photogrammetry Remote Sensing and Spatial InformationSciences vol 1ndash7 pp 341ndash346 2012
[17] H Nagaveena D Devaraj and S C Prasanna KumarldquoVessels segmentation in diabetic retinopathy by adaptivemedian thresholdingrdquo International Journal of Science andTechnology vol 1 no 1 pp 17ndash22 2013
[18] A Asuntha N Singh and A Srinivasan ldquoPSO genetic op-timization and SVM algorithm used for lung cancer
detectionrdquo Journal of Chemical and Pharmaceutical Researchvol 8 no 6 pp 351ndash359 2016
[19] P Bhuvaneswari and A Brintha+erese ldquoDetection of cancerin lung with K-NN classification using genetic algorithmrdquoInternational Conference on Nanomaterials and Technologiesvol 10 pp 433ndash440 2014
16 Computational and Mathematical Methods in Medicine
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom
72
77
82
87
92
97
Acc
urac
y
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 10 Comparative results of accuracy
0
5
10
15
20
25
False
pos
itive
rate
k-meansk-mediansPSO
IWPSOGCPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
Figure 8 Comparative results of the false positive rate value
0
5
10
15
20
25
False
neg
ativ
e rat
e
k-meansk-mediansPSO
Imag
e 1
Imag
e 2
Imag
e 3
Imag
e 4
Imag
e 5
Imag
e 6
Imag
e 7
Imag
e 8
Imag
e 9
Imag
e 10
Imag
e 11
Imag
e 12
Imag
e 13
Imag
e 14
Imag
e 15
Imag
e 16
Imag
e 17
Imag
e 18
Imag
e 19
Imag
e 20
Sample images
IWPSOGCPSO
Figure 9 Comparative results of the false negative rate value
8 Computational and Mathematical Methods in Medicine
Table 7 Statistical comparative result of accuracy
Images k-Means k-Median PSO IWPSO GCPSOImage 1 885937 893672 890624 889813 958079Image 2 882682 885622 889689 881810 954782Image 3 857501 862969 864850 861018 944204Image 4 842186 842695 849967 843584 936473Image 5 839216 845294 850299 846631 936792Image 6 825795 833654 835581 832876 930784Image 7 801519 807832 809438 807064 917434Image 8 790656 798186 802571 803544 915541Image 9 801606 809977 817340 814631 921453Image 10 790378 799611 806182 803943 915958Image 11 799755 806621 816421 812411 921061Image 12 815238 824514 831347 830334 928988Image 13 802806 812545 819351 818905 923198Image 14 793023 799876 809226 809025 918721Image 15 797600 807248 815586 816669 921649Image 16 792121 801291 810836 810896 919433Image 17 792229 803332 812824 810209 919531Image 18 807020 817792 818114 818081 923418Image 19 810816 818439 820954 820213 924662Image 20 801831 810973 811744 812033 919938
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 1
Image 2
Image 3
Image 4
Image 5
Image 6
Image 7
Image 8
Image 9
Image 10
(a)
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
(b)
Figure 11 Resultant images after preprocessing
Computational and Mathematical Methods in Medicine 9
eradicate the incidence of noise content and to improve theimage quality before an examination [4]+is part of work isknown as preprocessing In the preprocessing stage noiseremoval and contrast enhancement are two primary steps Inthe present study the performance results of medianadaptive median and average filters to isolate the presence ofspeckle noise have been compared +e coding for the samehas been implemented using MATLAB Furthermore theimage quality and visual appearance are improved byadaptive histogram equalization+e second stage of work issegmentation +is stage consists of applying five methodsnamely k-means k-median particle swarm optimization(PSO) inertia-weighted particle swarm optimization(IWPSO) and GCPSO +e tumor portion was extractedfrom the segmented results of the above-said five methodsand compared with manual extraction+e results show thatthe GCPSO-based segmentation has more accuracy than theothers Figure 1 depicts the process of operation for thepresent study
21 Median and Adaptive Median Filters +e median filterremoves the noise and retains the sharpness of the imageAccordance to the name each pixel is replaced by themedian value from the neighborhood pixels A 3 times 3 windowis used in this filter [5] +is is one of the best filters amongconventional filters which remove the speckle noise +esteps followed to construct the median filter are given inAlgorithm 1
Spatial processing to preserve the edge detail and toeliminate nonimpulsive noise by the adaptive median filterplays a vital role +e small structure in the image and edgesare retained by the adaptive median filter In the adaptivemedian filter the window size varies with respect to eachpixel
22 Average Filter +is is a simple filter which removesthe spatial noise from a digital image +e presence ofspatial noise is mainly due to the data acquisition process
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 12 Resultant images by k-means clustering
10 Computational and Mathematical Methods in Medicine
+e neighborhood mean value is measured for each andevery pixel and is replaced by the corresponding meanvalue +is process is repeated for every pixel in the image[5] All the pixels in the digital image are modified bysliding the operator over the entire range of pixels +esteps followed for the average filter are given inAlgorithm 2
23 Histogram Equalization Image enhancement is thetechnique which is used to improve the image quality Forbetter understanding and analysis it is mandatory to en-hance the contrast of medical images +e conventionalmethod used for this operation is histogram equalization Aminor adjustment on the intensity of image pixels is donein this method Each pixel is mapped to intensity pro-portional to its rank in the surrounding pixels +e stepsfollowed for histogram equalization are given in Algo-rithm 3 [6]
24 k-Means Clustering Algorithm +e simplest and con-ventional method in cluster analysis is the k-means clus-tering algorithm+is algorithm segregates the given datasetinto two or more clusters [7] +e accuracy of this methodcompletely depends on the selection of the cluster center Itis mandatory to select the optimum cluster center to get abetter result +e Euclidean distance is the general measureto segregate the dataset [8] Pixels are assigned to an indi-vidual cluster based on the Euclidean distance +e objectivefunction used in this algorithm is
J(v) 1113944C
i11113944
Ci
j1xi minus vj
1113874 11138752 (1)
where xi are the pixels vj are the cluster centers xi minus vj isthe Euclidean distance between xi and vjCi is the number ofdata points for the ith cluster and C is the number of clustercenters [9] +e steps followed for k-means clustering aregiven in Algorithm 4
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 13 Resultant images by k-median clustering
Computational and Mathematical Methods in Medicine 11
25 k-Median Clustering Algorithm +is is also a clusteringalgorithm slightly modified from the k-means algorithm Incentroid calculation instead of calculating the mean valuethe median value is considered +is algorithm significantlyreduces the error since there is no squared operation as inthe calculation of the Euclidean distance +e clustersformed by this method are more compact As an alternatethis approach uses the Lloyd-style iteration +e steps fol-lowed for k-median clustering are given in Algorithm 5 [10]
26 Particle Swarm Optimization PSO is a metaheuristicalgorithm used efficiently in medical image analysis [11] Itmimics the social behavior of the birds searching for food [12]+e fundamental idea of PSO is sharing and communicatingthe information In this approach each particle has initialposition and velocity Based on the fitness value the velocity
and position are updated +e relevant two equations in PSOto update the position and velocity are as follows [11 12]
v(t + 1) v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minus x(t)]
x(t + 1) x(t) + v(t + 1)
(2)
where r1 and r2 are the random numbers and the accel-eration coefficients c1 and c2 are two positive constants+e success of PSO relies on the fitness function +efollowing fitness function has been used for the presentstudy
maximizef 1113944n
i1
intercluster distanceintracluster distance
(3)
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 14 Resultant images by the PSO algorithm
12 Computational and Mathematical Methods in Medicine
where n is the number of clusters +e steps followed for theparticle swarm optimization are shown in Algorithm 6
27 Inertia-Weighted Particle Swarm Optimization +eexploration and exploitation in PSO are based on the inertiaweight +e basic PSO presented by Eberhart and Kennedyin 1995 has no inertia weight In 1998 Shi and Eberhartintroduced the concept of inertia weight by adding constantinertia weight +ey stated that a significant inertia weightfacilitates a global search while a small inertia weight fa-cilitates a local search [14] +is enhances the convergencerate and reduces the number of iterations Inertia weight lessthan 1 in general improves the results +e used methodimproves the convergence rate and saves the time taken andsome iterations
+e resulting velocity update equation becomes
v(t + 1) wlowast v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minusx(t)](4)
where w is the inertia weight with constant inertia weight w
07 and random inertia weight w 05 + rand()2
28 Guaranteed Convergence Particle Swarm Optimization+e GCPSO focuses on a new particle which deals with thecurrent best position in the region In this task this particleis treated as a member of the swarm and the velocity updateequation for this new particle is given as follows [15]
vφ(t + 1) xφ(t) + pbest(t) + ωvφ(t) + ρ(t)(1minus 2r) (5)
+e search ability is increased by the social part +is willimprove the random search in the area around the gbestposition +e random vector and diameter of the search area
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 15 Resultant images by IWPSO algorithm clustering
Computational and Mathematical Methods in Medicine 13
are r and ρ(t) respectively +e range of the random vectorlies between 0 and 1 +e diameter of the search area can beupdated using the following equation
ρ(t + 1)
2ρ(t) successesgt sc
115
1113874 1113875ρ(t) failuresgt fc
ρ(t) otherwise
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎩
(6)
where the terms successes and failures are defined asthe number of consecutive successes and failures re-spectively +e threshold parameters sc and fc are de-termined empirically Since it is hard to obtain a bettervalue in only a few iterations in a high-dimensional searchspace the recommended values are thus sc 15 and fc 5
On some benchmark tests the GCPSO has shown anexcellent performance of locating the minimal of a spaceafter unimodal with only a small amount of particles +esteps to be followed for the GCPSO are shown inAlgorithm 7
3 Performance Measures
Certain performance measures are used to evaluate theresults obtained from medical image segmentation +e listof performance measures used to assess the filter operation isshown in Figure 2 [16] Let If be the image after noise re-duction and I0 be the noisy image
Performance measures used for the evaluation of theresults of the segmentation algorithm are given in Figure 3[17]
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 16 Resultant images by GCPSO algorithm clustering
14 Computational and Mathematical Methods in Medicine
4 Results and Discussion
+e used methods are practically implemented usingMATLAB coding and the results were verified
In the preprocessing stage a comparison was donebetween the performance of median adaptive median andmean filters +e SSI and SMPI values are shown in Table 1and Figures 4 and 5 From the results it is evident that theadaptive median filter has accurate characteristics than themean and median filters for medical image segmentation
+e segmentation accuracy was measured using the truepositive rate true negative rate false positive rate and falsenegative rate by comparing the results from the algorithmwithmanual segmentation results +e practical results of the k-means clustering segmentation algorithm are shown in Table 2
+e practical results of the k-median clustering seg-mentation algorithm are shown in Table 3
+e practical results of the PSO-based segmentationalgorithm are shown in Table 4
+e practical results of the IWPSO segmentation algo-rithm are shown in Table 5
+e practical results of the GCPSO segmentation algo-rithm are shown in Table 6
+e graphical view of the comparison of the true positiverate true negative rate false positive rate and false negativerate for the algorithms used is shown in Figures 6ndash9 It isproved that the true positive and true negative rates are highand false positive and false negative rates are low for theGCPSO algorithm
+e comparative evaluation based on the accuracy of thesegmentation is shown in Table 7 and Figure 10 +e resultsindicate that the GCPSO-based technique has the highestaverage value of accuracy than the other methods
+e resultant images after preprocessing are shown inFigures 11(a) and 11(b)
+e resultant images after segmentation using k-meansclustering are shown in Figure 12
+e resultant images after segmentation using k-medianclustering are shown in Figure 13
+e resultant images after segmentation using the PSOalgorithm are shown in Figure 14
+e resultant images after segmentation using theIWPSO algorithm are shown in Figure 15
+e resultant images after segmentation using theGCPSO algorithm are shown in Figure 16
In an earlier research lung cancer detection was doneusing PSO genetic optimization and SVM algorithm withthe Gabor filter and produced an accuracy of 895 [18]+emethod to detect lung cancer by means of K-NN classifi-cation using the genetic algorithm produced a maximumaccuracy of 90 [19]+e comparative results with respect tothe above-said methods are shown in Table 8
+e graphical comparative analysis between the used andexisting methods is shown in Figure 17
5 Conclusion
In this study various optimization algorithms have beenevaluated to detect the tumor Medical images often need
preprocessing before being subjected to statistical analysis+e adaptive median filter has better results than medianand mean filters because the speckle suppression index andspeckle and mean preservation index values are lower for theadaptive median filter Comparing the five algorithms theaccuracy of the tumor extraction is improved in GCPSOwith the highest accuracy of 958079 and it obtained above90 of precision in all the 20 images It is more accuratewhen compared to the previous method which had an ac-curacy of 90 in 4 out of 10 datasets only In future studiesthe use of more number of optimization algorithms will beincluded to improve the accuracy
Data Availability
+eCTimages data used to support the findings of this studyhave been deposited in the LungCT-Diagnosis repository(doiorg107937K9TCIA2015A6V7JIWX)
Conflicts of Interest
+e authors declare that there are no conflicts of interestregarding the publication of this paper
References
[1] A A Brindha S Indirani and A Srinivasan ldquoLung cancerdetection using SVM algorithm and optimization techniquesrdquoJournal of Chemical and Pharmaceutical Sciences vol 9 no 42016
[2] M Kurkure and A +akare ldquoIntroducing automated systemfor lung cancer detection using Evolutionary ApproachrdquoInternational Journal of Engineering and Computer Sciencevol 5 no 5 pp 16736ndash16739 2016
[3] B Rani A K Goel and R Kaur ldquoA modified approach forlung cancer detection using bacterial forging optimization
868890929496
PSO GA and SVM algorithm
K-NNclassification
using GA
Proposed GCPSOmethod
8950 90
9581
Acc
urac
y
Various algorithms
Comparison between existing and projected methods
Figure 17 Graphical view of accuracy
Table 8 Comparative analysis of accuracy of the projected methodwith various methods
Various methods Accuracy ()PSO GA and SVM algorithm [18] 8950K-NN classification using GA [19] 9000Projected GCPSO method 9581
Computational and Mathematical Methods in Medicine 15
algorithmrdquo International Journal of Scientific Research En-gineering and Technology vol 5 no 1 2016
[4] N Panpaliya N Tadas S Bobade R Aglawe andA Gudadhe ldquoA survey on early detection and prediction oflung cancerrdquo International Journal of Computer Science andMobile Computing vol 4 no 1 pp 175ndash184 2015
[5] G Gupta ldquoAlgorithm for image processing using improvedmedian filter and comparison of mean median and improvedmedian filterrdquo International Journal of Soft Computing andEngineering vol 1 no 5 2011
[6] Tutorial point with digital image processing httpwwwtutorialspointcomdiphistogram_equalizationhtm
[7] K Venkatalakshmi and S M Shalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and K-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Melbourne VIC Australia December2005
[8] K Venkatalakshmi and S M Shalinie ldquoMultispectral imageclassification using modified k-means clusteringrdquo In-ternational Journal on Neural and Mass-Parallel Computingand Information Systems vol 17 no 2 pp 113ndash120 2007
[9] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercy Shalinie ldquoMultispectral image clustering using en-hanced genetic k-means algorithmrdquo Information TechnologyJournal vol 6 no 4 pp 554ndash560 2007
[10] P I Dalatu ldquoTime complexity of k-means and k-mediansclustering algorithms in outliers detectionrdquo Global Journal ofPure and Applied Mathematics vol 12 no 5 pp 4405ndash44182016
[11] K Senthilkumar K Venkatalakshmi K Karthikeyan andP Kathirkamasundari ldquoAn efficient method for segmentingdigital image using a hybrid model of particle swarm opti-mization and artificial bee colony algorithmrdquo InternationalJournal of Applied Engineering Research vol 10 pp 444ndash4492015
[12] K Venkatalakshmi and S Mercyshalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and k-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Bangalore India December 2005
[13] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercyshalinie ldquoA customized particle swarm optimizationfor classification of multispectral imagery based on featurefusionrdquo International Arab Journal of Information Technologyvol 5 no 4 pp 327ndash333 2008
[14] J C Bansal and P K Singh ldquoInertia weight strategies inparticle swarm optimizationrdquo in Proceedings of IEEE WorldCongress on Nature and Biologically Inspired Computing pp640ndash647 Salamanca Spain October 2011
[15] P K Patel V Sharma and K Gupta ldquoGuaranteed conver-gence particle swarm optimization using Personal Bestrdquo In-ternational Journal of Computer Applications vol 73 no 7pp 6ndash10 2013
[16] X Wang L Ge and X Li ldquoEvaluation of filters for Envi-satAsar speckle suppression in pasture areardquo ISPRS Annals ofthe Photogrammetry Remote Sensing and Spatial InformationSciences vol 1ndash7 pp 341ndash346 2012
[17] H Nagaveena D Devaraj and S C Prasanna KumarldquoVessels segmentation in diabetic retinopathy by adaptivemedian thresholdingrdquo International Journal of Science andTechnology vol 1 no 1 pp 17ndash22 2013
[18] A Asuntha N Singh and A Srinivasan ldquoPSO genetic op-timization and SVM algorithm used for lung cancer
detectionrdquo Journal of Chemical and Pharmaceutical Researchvol 8 no 6 pp 351ndash359 2016
[19] P Bhuvaneswari and A Brintha+erese ldquoDetection of cancerin lung with K-NN classification using genetic algorithmrdquoInternational Conference on Nanomaterials and Technologiesvol 10 pp 433ndash440 2014
16 Computational and Mathematical Methods in Medicine
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom
Table 7 Statistical comparative result of accuracy
Images k-Means k-Median PSO IWPSO GCPSOImage 1 885937 893672 890624 889813 958079Image 2 882682 885622 889689 881810 954782Image 3 857501 862969 864850 861018 944204Image 4 842186 842695 849967 843584 936473Image 5 839216 845294 850299 846631 936792Image 6 825795 833654 835581 832876 930784Image 7 801519 807832 809438 807064 917434Image 8 790656 798186 802571 803544 915541Image 9 801606 809977 817340 814631 921453Image 10 790378 799611 806182 803943 915958Image 11 799755 806621 816421 812411 921061Image 12 815238 824514 831347 830334 928988Image 13 802806 812545 819351 818905 923198Image 14 793023 799876 809226 809025 918721Image 15 797600 807248 815586 816669 921649Image 16 792121 801291 810836 810896 919433Image 17 792229 803332 812824 810209 919531Image 18 807020 817792 818114 818081 923418Image 19 810816 818439 820954 820213 924662Image 20 801831 810973 811744 812033 919938
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 1
Image 2
Image 3
Image 4
Image 5
Image 6
Image 7
Image 8
Image 9
Image 10
(a)
Inputimage
Averagefilter
output
Medianfilter
output
Adaptivemedian
filter output
Contrastenhancement
outputSampleimages
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
(b)
Figure 11 Resultant images after preprocessing
Computational and Mathematical Methods in Medicine 9
eradicate the incidence of noise content and to improve theimage quality before an examination [4]+is part of work isknown as preprocessing In the preprocessing stage noiseremoval and contrast enhancement are two primary steps Inthe present study the performance results of medianadaptive median and average filters to isolate the presence ofspeckle noise have been compared +e coding for the samehas been implemented using MATLAB Furthermore theimage quality and visual appearance are improved byadaptive histogram equalization+e second stage of work issegmentation +is stage consists of applying five methodsnamely k-means k-median particle swarm optimization(PSO) inertia-weighted particle swarm optimization(IWPSO) and GCPSO +e tumor portion was extractedfrom the segmented results of the above-said five methodsand compared with manual extraction+e results show thatthe GCPSO-based segmentation has more accuracy than theothers Figure 1 depicts the process of operation for thepresent study
21 Median and Adaptive Median Filters +e median filterremoves the noise and retains the sharpness of the imageAccordance to the name each pixel is replaced by themedian value from the neighborhood pixels A 3 times 3 windowis used in this filter [5] +is is one of the best filters amongconventional filters which remove the speckle noise +esteps followed to construct the median filter are given inAlgorithm 1
Spatial processing to preserve the edge detail and toeliminate nonimpulsive noise by the adaptive median filterplays a vital role +e small structure in the image and edgesare retained by the adaptive median filter In the adaptivemedian filter the window size varies with respect to eachpixel
22 Average Filter +is is a simple filter which removesthe spatial noise from a digital image +e presence ofspatial noise is mainly due to the data acquisition process
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 12 Resultant images by k-means clustering
10 Computational and Mathematical Methods in Medicine
+e neighborhood mean value is measured for each andevery pixel and is replaced by the corresponding meanvalue +is process is repeated for every pixel in the image[5] All the pixels in the digital image are modified bysliding the operator over the entire range of pixels +esteps followed for the average filter are given inAlgorithm 2
23 Histogram Equalization Image enhancement is thetechnique which is used to improve the image quality Forbetter understanding and analysis it is mandatory to en-hance the contrast of medical images +e conventionalmethod used for this operation is histogram equalization Aminor adjustment on the intensity of image pixels is donein this method Each pixel is mapped to intensity pro-portional to its rank in the surrounding pixels +e stepsfollowed for histogram equalization are given in Algo-rithm 3 [6]
24 k-Means Clustering Algorithm +e simplest and con-ventional method in cluster analysis is the k-means clus-tering algorithm+is algorithm segregates the given datasetinto two or more clusters [7] +e accuracy of this methodcompletely depends on the selection of the cluster center Itis mandatory to select the optimum cluster center to get abetter result +e Euclidean distance is the general measureto segregate the dataset [8] Pixels are assigned to an indi-vidual cluster based on the Euclidean distance +e objectivefunction used in this algorithm is
J(v) 1113944C
i11113944
Ci
j1xi minus vj
1113874 11138752 (1)
where xi are the pixels vj are the cluster centers xi minus vj isthe Euclidean distance between xi and vjCi is the number ofdata points for the ith cluster and C is the number of clustercenters [9] +e steps followed for k-means clustering aregiven in Algorithm 4
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 13 Resultant images by k-median clustering
Computational and Mathematical Methods in Medicine 11
25 k-Median Clustering Algorithm +is is also a clusteringalgorithm slightly modified from the k-means algorithm Incentroid calculation instead of calculating the mean valuethe median value is considered +is algorithm significantlyreduces the error since there is no squared operation as inthe calculation of the Euclidean distance +e clustersformed by this method are more compact As an alternatethis approach uses the Lloyd-style iteration +e steps fol-lowed for k-median clustering are given in Algorithm 5 [10]
26 Particle Swarm Optimization PSO is a metaheuristicalgorithm used efficiently in medical image analysis [11] Itmimics the social behavior of the birds searching for food [12]+e fundamental idea of PSO is sharing and communicatingthe information In this approach each particle has initialposition and velocity Based on the fitness value the velocity
and position are updated +e relevant two equations in PSOto update the position and velocity are as follows [11 12]
v(t + 1) v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minus x(t)]
x(t + 1) x(t) + v(t + 1)
(2)
where r1 and r2 are the random numbers and the accel-eration coefficients c1 and c2 are two positive constants+e success of PSO relies on the fitness function +efollowing fitness function has been used for the presentstudy
maximizef 1113944n
i1
intercluster distanceintracluster distance
(3)
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 14 Resultant images by the PSO algorithm
12 Computational and Mathematical Methods in Medicine
where n is the number of clusters +e steps followed for theparticle swarm optimization are shown in Algorithm 6
27 Inertia-Weighted Particle Swarm Optimization +eexploration and exploitation in PSO are based on the inertiaweight +e basic PSO presented by Eberhart and Kennedyin 1995 has no inertia weight In 1998 Shi and Eberhartintroduced the concept of inertia weight by adding constantinertia weight +ey stated that a significant inertia weightfacilitates a global search while a small inertia weight fa-cilitates a local search [14] +is enhances the convergencerate and reduces the number of iterations Inertia weight lessthan 1 in general improves the results +e used methodimproves the convergence rate and saves the time taken andsome iterations
+e resulting velocity update equation becomes
v(t + 1) wlowast v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minusx(t)](4)
where w is the inertia weight with constant inertia weight w
07 and random inertia weight w 05 + rand()2
28 Guaranteed Convergence Particle Swarm Optimization+e GCPSO focuses on a new particle which deals with thecurrent best position in the region In this task this particleis treated as a member of the swarm and the velocity updateequation for this new particle is given as follows [15]
vφ(t + 1) xφ(t) + pbest(t) + ωvφ(t) + ρ(t)(1minus 2r) (5)
+e search ability is increased by the social part +is willimprove the random search in the area around the gbestposition +e random vector and diameter of the search area
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 15 Resultant images by IWPSO algorithm clustering
Computational and Mathematical Methods in Medicine 13
are r and ρ(t) respectively +e range of the random vectorlies between 0 and 1 +e diameter of the search area can beupdated using the following equation
ρ(t + 1)
2ρ(t) successesgt sc
115
1113874 1113875ρ(t) failuresgt fc
ρ(t) otherwise
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎩
(6)
where the terms successes and failures are defined asthe number of consecutive successes and failures re-spectively +e threshold parameters sc and fc are de-termined empirically Since it is hard to obtain a bettervalue in only a few iterations in a high-dimensional searchspace the recommended values are thus sc 15 and fc 5
On some benchmark tests the GCPSO has shown anexcellent performance of locating the minimal of a spaceafter unimodal with only a small amount of particles +esteps to be followed for the GCPSO are shown inAlgorithm 7
3 Performance Measures
Certain performance measures are used to evaluate theresults obtained from medical image segmentation +e listof performance measures used to assess the filter operation isshown in Figure 2 [16] Let If be the image after noise re-duction and I0 be the noisy image
Performance measures used for the evaluation of theresults of the segmentation algorithm are given in Figure 3[17]
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 16 Resultant images by GCPSO algorithm clustering
14 Computational and Mathematical Methods in Medicine
4 Results and Discussion
+e used methods are practically implemented usingMATLAB coding and the results were verified
In the preprocessing stage a comparison was donebetween the performance of median adaptive median andmean filters +e SSI and SMPI values are shown in Table 1and Figures 4 and 5 From the results it is evident that theadaptive median filter has accurate characteristics than themean and median filters for medical image segmentation
+e segmentation accuracy was measured using the truepositive rate true negative rate false positive rate and falsenegative rate by comparing the results from the algorithmwithmanual segmentation results +e practical results of the k-means clustering segmentation algorithm are shown in Table 2
+e practical results of the k-median clustering seg-mentation algorithm are shown in Table 3
+e practical results of the PSO-based segmentationalgorithm are shown in Table 4
+e practical results of the IWPSO segmentation algo-rithm are shown in Table 5
+e practical results of the GCPSO segmentation algo-rithm are shown in Table 6
+e graphical view of the comparison of the true positiverate true negative rate false positive rate and false negativerate for the algorithms used is shown in Figures 6ndash9 It isproved that the true positive and true negative rates are highand false positive and false negative rates are low for theGCPSO algorithm
+e comparative evaluation based on the accuracy of thesegmentation is shown in Table 7 and Figure 10 +e resultsindicate that the GCPSO-based technique has the highestaverage value of accuracy than the other methods
+e resultant images after preprocessing are shown inFigures 11(a) and 11(b)
+e resultant images after segmentation using k-meansclustering are shown in Figure 12
+e resultant images after segmentation using k-medianclustering are shown in Figure 13
+e resultant images after segmentation using the PSOalgorithm are shown in Figure 14
+e resultant images after segmentation using theIWPSO algorithm are shown in Figure 15
+e resultant images after segmentation using theGCPSO algorithm are shown in Figure 16
In an earlier research lung cancer detection was doneusing PSO genetic optimization and SVM algorithm withthe Gabor filter and produced an accuracy of 895 [18]+emethod to detect lung cancer by means of K-NN classifi-cation using the genetic algorithm produced a maximumaccuracy of 90 [19]+e comparative results with respect tothe above-said methods are shown in Table 8
+e graphical comparative analysis between the used andexisting methods is shown in Figure 17
5 Conclusion
In this study various optimization algorithms have beenevaluated to detect the tumor Medical images often need
preprocessing before being subjected to statistical analysis+e adaptive median filter has better results than medianand mean filters because the speckle suppression index andspeckle and mean preservation index values are lower for theadaptive median filter Comparing the five algorithms theaccuracy of the tumor extraction is improved in GCPSOwith the highest accuracy of 958079 and it obtained above90 of precision in all the 20 images It is more accuratewhen compared to the previous method which had an ac-curacy of 90 in 4 out of 10 datasets only In future studiesthe use of more number of optimization algorithms will beincluded to improve the accuracy
Data Availability
+eCTimages data used to support the findings of this studyhave been deposited in the LungCT-Diagnosis repository(doiorg107937K9TCIA2015A6V7JIWX)
Conflicts of Interest
+e authors declare that there are no conflicts of interestregarding the publication of this paper
References
[1] A A Brindha S Indirani and A Srinivasan ldquoLung cancerdetection using SVM algorithm and optimization techniquesrdquoJournal of Chemical and Pharmaceutical Sciences vol 9 no 42016
[2] M Kurkure and A +akare ldquoIntroducing automated systemfor lung cancer detection using Evolutionary ApproachrdquoInternational Journal of Engineering and Computer Sciencevol 5 no 5 pp 16736ndash16739 2016
[3] B Rani A K Goel and R Kaur ldquoA modified approach forlung cancer detection using bacterial forging optimization
868890929496
PSO GA and SVM algorithm
K-NNclassification
using GA
Proposed GCPSOmethod
8950 90
9581
Acc
urac
y
Various algorithms
Comparison between existing and projected methods
Figure 17 Graphical view of accuracy
Table 8 Comparative analysis of accuracy of the projected methodwith various methods
Various methods Accuracy ()PSO GA and SVM algorithm [18] 8950K-NN classification using GA [19] 9000Projected GCPSO method 9581
Computational and Mathematical Methods in Medicine 15
algorithmrdquo International Journal of Scientific Research En-gineering and Technology vol 5 no 1 2016
[4] N Panpaliya N Tadas S Bobade R Aglawe andA Gudadhe ldquoA survey on early detection and prediction oflung cancerrdquo International Journal of Computer Science andMobile Computing vol 4 no 1 pp 175ndash184 2015
[5] G Gupta ldquoAlgorithm for image processing using improvedmedian filter and comparison of mean median and improvedmedian filterrdquo International Journal of Soft Computing andEngineering vol 1 no 5 2011
[6] Tutorial point with digital image processing httpwwwtutorialspointcomdiphistogram_equalizationhtm
[7] K Venkatalakshmi and S M Shalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and K-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Melbourne VIC Australia December2005
[8] K Venkatalakshmi and S M Shalinie ldquoMultispectral imageclassification using modified k-means clusteringrdquo In-ternational Journal on Neural and Mass-Parallel Computingand Information Systems vol 17 no 2 pp 113ndash120 2007
[9] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercy Shalinie ldquoMultispectral image clustering using en-hanced genetic k-means algorithmrdquo Information TechnologyJournal vol 6 no 4 pp 554ndash560 2007
[10] P I Dalatu ldquoTime complexity of k-means and k-mediansclustering algorithms in outliers detectionrdquo Global Journal ofPure and Applied Mathematics vol 12 no 5 pp 4405ndash44182016
[11] K Senthilkumar K Venkatalakshmi K Karthikeyan andP Kathirkamasundari ldquoAn efficient method for segmentingdigital image using a hybrid model of particle swarm opti-mization and artificial bee colony algorithmrdquo InternationalJournal of Applied Engineering Research vol 10 pp 444ndash4492015
[12] K Venkatalakshmi and S Mercyshalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and k-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Bangalore India December 2005
[13] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercyshalinie ldquoA customized particle swarm optimizationfor classification of multispectral imagery based on featurefusionrdquo International Arab Journal of Information Technologyvol 5 no 4 pp 327ndash333 2008
[14] J C Bansal and P K Singh ldquoInertia weight strategies inparticle swarm optimizationrdquo in Proceedings of IEEE WorldCongress on Nature and Biologically Inspired Computing pp640ndash647 Salamanca Spain October 2011
[15] P K Patel V Sharma and K Gupta ldquoGuaranteed conver-gence particle swarm optimization using Personal Bestrdquo In-ternational Journal of Computer Applications vol 73 no 7pp 6ndash10 2013
[16] X Wang L Ge and X Li ldquoEvaluation of filters for Envi-satAsar speckle suppression in pasture areardquo ISPRS Annals ofthe Photogrammetry Remote Sensing and Spatial InformationSciences vol 1ndash7 pp 341ndash346 2012
[17] H Nagaveena D Devaraj and S C Prasanna KumarldquoVessels segmentation in diabetic retinopathy by adaptivemedian thresholdingrdquo International Journal of Science andTechnology vol 1 no 1 pp 17ndash22 2013
[18] A Asuntha N Singh and A Srinivasan ldquoPSO genetic op-timization and SVM algorithm used for lung cancer
detectionrdquo Journal of Chemical and Pharmaceutical Researchvol 8 no 6 pp 351ndash359 2016
[19] P Bhuvaneswari and A Brintha+erese ldquoDetection of cancerin lung with K-NN classification using genetic algorithmrdquoInternational Conference on Nanomaterials and Technologiesvol 10 pp 433ndash440 2014
16 Computational and Mathematical Methods in Medicine
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom
eradicate the incidence of noise content and to improve theimage quality before an examination [4]+is part of work isknown as preprocessing In the preprocessing stage noiseremoval and contrast enhancement are two primary steps Inthe present study the performance results of medianadaptive median and average filters to isolate the presence ofspeckle noise have been compared +e coding for the samehas been implemented using MATLAB Furthermore theimage quality and visual appearance are improved byadaptive histogram equalization+e second stage of work issegmentation +is stage consists of applying five methodsnamely k-means k-median particle swarm optimization(PSO) inertia-weighted particle swarm optimization(IWPSO) and GCPSO +e tumor portion was extractedfrom the segmented results of the above-said five methodsand compared with manual extraction+e results show thatthe GCPSO-based segmentation has more accuracy than theothers Figure 1 depicts the process of operation for thepresent study
21 Median and Adaptive Median Filters +e median filterremoves the noise and retains the sharpness of the imageAccordance to the name each pixel is replaced by themedian value from the neighborhood pixels A 3 times 3 windowis used in this filter [5] +is is one of the best filters amongconventional filters which remove the speckle noise +esteps followed to construct the median filter are given inAlgorithm 1
Spatial processing to preserve the edge detail and toeliminate nonimpulsive noise by the adaptive median filterplays a vital role +e small structure in the image and edgesare retained by the adaptive median filter In the adaptivemedian filter the window size varies with respect to eachpixel
22 Average Filter +is is a simple filter which removesthe spatial noise from a digital image +e presence ofspatial noise is mainly due to the data acquisition process
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 12 Resultant images by k-means clustering
10 Computational and Mathematical Methods in Medicine
+e neighborhood mean value is measured for each andevery pixel and is replaced by the corresponding meanvalue +is process is repeated for every pixel in the image[5] All the pixels in the digital image are modified bysliding the operator over the entire range of pixels +esteps followed for the average filter are given inAlgorithm 2
23 Histogram Equalization Image enhancement is thetechnique which is used to improve the image quality Forbetter understanding and analysis it is mandatory to en-hance the contrast of medical images +e conventionalmethod used for this operation is histogram equalization Aminor adjustment on the intensity of image pixels is donein this method Each pixel is mapped to intensity pro-portional to its rank in the surrounding pixels +e stepsfollowed for histogram equalization are given in Algo-rithm 3 [6]
24 k-Means Clustering Algorithm +e simplest and con-ventional method in cluster analysis is the k-means clus-tering algorithm+is algorithm segregates the given datasetinto two or more clusters [7] +e accuracy of this methodcompletely depends on the selection of the cluster center Itis mandatory to select the optimum cluster center to get abetter result +e Euclidean distance is the general measureto segregate the dataset [8] Pixels are assigned to an indi-vidual cluster based on the Euclidean distance +e objectivefunction used in this algorithm is
J(v) 1113944C
i11113944
Ci
j1xi minus vj
1113874 11138752 (1)
where xi are the pixels vj are the cluster centers xi minus vj isthe Euclidean distance between xi and vjCi is the number ofdata points for the ith cluster and C is the number of clustercenters [9] +e steps followed for k-means clustering aregiven in Algorithm 4
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 13 Resultant images by k-median clustering
Computational and Mathematical Methods in Medicine 11
25 k-Median Clustering Algorithm +is is also a clusteringalgorithm slightly modified from the k-means algorithm Incentroid calculation instead of calculating the mean valuethe median value is considered +is algorithm significantlyreduces the error since there is no squared operation as inthe calculation of the Euclidean distance +e clustersformed by this method are more compact As an alternatethis approach uses the Lloyd-style iteration +e steps fol-lowed for k-median clustering are given in Algorithm 5 [10]
26 Particle Swarm Optimization PSO is a metaheuristicalgorithm used efficiently in medical image analysis [11] Itmimics the social behavior of the birds searching for food [12]+e fundamental idea of PSO is sharing and communicatingthe information In this approach each particle has initialposition and velocity Based on the fitness value the velocity
and position are updated +e relevant two equations in PSOto update the position and velocity are as follows [11 12]
v(t + 1) v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minus x(t)]
x(t + 1) x(t) + v(t + 1)
(2)
where r1 and r2 are the random numbers and the accel-eration coefficients c1 and c2 are two positive constants+e success of PSO relies on the fitness function +efollowing fitness function has been used for the presentstudy
maximizef 1113944n
i1
intercluster distanceintracluster distance
(3)
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 14 Resultant images by the PSO algorithm
12 Computational and Mathematical Methods in Medicine
where n is the number of clusters +e steps followed for theparticle swarm optimization are shown in Algorithm 6
27 Inertia-Weighted Particle Swarm Optimization +eexploration and exploitation in PSO are based on the inertiaweight +e basic PSO presented by Eberhart and Kennedyin 1995 has no inertia weight In 1998 Shi and Eberhartintroduced the concept of inertia weight by adding constantinertia weight +ey stated that a significant inertia weightfacilitates a global search while a small inertia weight fa-cilitates a local search [14] +is enhances the convergencerate and reduces the number of iterations Inertia weight lessthan 1 in general improves the results +e used methodimproves the convergence rate and saves the time taken andsome iterations
+e resulting velocity update equation becomes
v(t + 1) wlowast v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minusx(t)](4)
where w is the inertia weight with constant inertia weight w
07 and random inertia weight w 05 + rand()2
28 Guaranteed Convergence Particle Swarm Optimization+e GCPSO focuses on a new particle which deals with thecurrent best position in the region In this task this particleis treated as a member of the swarm and the velocity updateequation for this new particle is given as follows [15]
vφ(t + 1) xφ(t) + pbest(t) + ωvφ(t) + ρ(t)(1minus 2r) (5)
+e search ability is increased by the social part +is willimprove the random search in the area around the gbestposition +e random vector and diameter of the search area
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 15 Resultant images by IWPSO algorithm clustering
Computational and Mathematical Methods in Medicine 13
are r and ρ(t) respectively +e range of the random vectorlies between 0 and 1 +e diameter of the search area can beupdated using the following equation
ρ(t + 1)
2ρ(t) successesgt sc
115
1113874 1113875ρ(t) failuresgt fc
ρ(t) otherwise
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎩
(6)
where the terms successes and failures are defined asthe number of consecutive successes and failures re-spectively +e threshold parameters sc and fc are de-termined empirically Since it is hard to obtain a bettervalue in only a few iterations in a high-dimensional searchspace the recommended values are thus sc 15 and fc 5
On some benchmark tests the GCPSO has shown anexcellent performance of locating the minimal of a spaceafter unimodal with only a small amount of particles +esteps to be followed for the GCPSO are shown inAlgorithm 7
3 Performance Measures
Certain performance measures are used to evaluate theresults obtained from medical image segmentation +e listof performance measures used to assess the filter operation isshown in Figure 2 [16] Let If be the image after noise re-duction and I0 be the noisy image
Performance measures used for the evaluation of theresults of the segmentation algorithm are given in Figure 3[17]
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 16 Resultant images by GCPSO algorithm clustering
14 Computational and Mathematical Methods in Medicine
4 Results and Discussion
+e used methods are practically implemented usingMATLAB coding and the results were verified
In the preprocessing stage a comparison was donebetween the performance of median adaptive median andmean filters +e SSI and SMPI values are shown in Table 1and Figures 4 and 5 From the results it is evident that theadaptive median filter has accurate characteristics than themean and median filters for medical image segmentation
+e segmentation accuracy was measured using the truepositive rate true negative rate false positive rate and falsenegative rate by comparing the results from the algorithmwithmanual segmentation results +e practical results of the k-means clustering segmentation algorithm are shown in Table 2
+e practical results of the k-median clustering seg-mentation algorithm are shown in Table 3
+e practical results of the PSO-based segmentationalgorithm are shown in Table 4
+e practical results of the IWPSO segmentation algo-rithm are shown in Table 5
+e practical results of the GCPSO segmentation algo-rithm are shown in Table 6
+e graphical view of the comparison of the true positiverate true negative rate false positive rate and false negativerate for the algorithms used is shown in Figures 6ndash9 It isproved that the true positive and true negative rates are highand false positive and false negative rates are low for theGCPSO algorithm
+e comparative evaluation based on the accuracy of thesegmentation is shown in Table 7 and Figure 10 +e resultsindicate that the GCPSO-based technique has the highestaverage value of accuracy than the other methods
+e resultant images after preprocessing are shown inFigures 11(a) and 11(b)
+e resultant images after segmentation using k-meansclustering are shown in Figure 12
+e resultant images after segmentation using k-medianclustering are shown in Figure 13
+e resultant images after segmentation using the PSOalgorithm are shown in Figure 14
+e resultant images after segmentation using theIWPSO algorithm are shown in Figure 15
+e resultant images after segmentation using theGCPSO algorithm are shown in Figure 16
In an earlier research lung cancer detection was doneusing PSO genetic optimization and SVM algorithm withthe Gabor filter and produced an accuracy of 895 [18]+emethod to detect lung cancer by means of K-NN classifi-cation using the genetic algorithm produced a maximumaccuracy of 90 [19]+e comparative results with respect tothe above-said methods are shown in Table 8
+e graphical comparative analysis between the used andexisting methods is shown in Figure 17
5 Conclusion
In this study various optimization algorithms have beenevaluated to detect the tumor Medical images often need
preprocessing before being subjected to statistical analysis+e adaptive median filter has better results than medianand mean filters because the speckle suppression index andspeckle and mean preservation index values are lower for theadaptive median filter Comparing the five algorithms theaccuracy of the tumor extraction is improved in GCPSOwith the highest accuracy of 958079 and it obtained above90 of precision in all the 20 images It is more accuratewhen compared to the previous method which had an ac-curacy of 90 in 4 out of 10 datasets only In future studiesthe use of more number of optimization algorithms will beincluded to improve the accuracy
Data Availability
+eCTimages data used to support the findings of this studyhave been deposited in the LungCT-Diagnosis repository(doiorg107937K9TCIA2015A6V7JIWX)
Conflicts of Interest
+e authors declare that there are no conflicts of interestregarding the publication of this paper
References
[1] A A Brindha S Indirani and A Srinivasan ldquoLung cancerdetection using SVM algorithm and optimization techniquesrdquoJournal of Chemical and Pharmaceutical Sciences vol 9 no 42016
[2] M Kurkure and A +akare ldquoIntroducing automated systemfor lung cancer detection using Evolutionary ApproachrdquoInternational Journal of Engineering and Computer Sciencevol 5 no 5 pp 16736ndash16739 2016
[3] B Rani A K Goel and R Kaur ldquoA modified approach forlung cancer detection using bacterial forging optimization
868890929496
PSO GA and SVM algorithm
K-NNclassification
using GA
Proposed GCPSOmethod
8950 90
9581
Acc
urac
y
Various algorithms
Comparison between existing and projected methods
Figure 17 Graphical view of accuracy
Table 8 Comparative analysis of accuracy of the projected methodwith various methods
Various methods Accuracy ()PSO GA and SVM algorithm [18] 8950K-NN classification using GA [19] 9000Projected GCPSO method 9581
Computational and Mathematical Methods in Medicine 15
algorithmrdquo International Journal of Scientific Research En-gineering and Technology vol 5 no 1 2016
[4] N Panpaliya N Tadas S Bobade R Aglawe andA Gudadhe ldquoA survey on early detection and prediction oflung cancerrdquo International Journal of Computer Science andMobile Computing vol 4 no 1 pp 175ndash184 2015
[5] G Gupta ldquoAlgorithm for image processing using improvedmedian filter and comparison of mean median and improvedmedian filterrdquo International Journal of Soft Computing andEngineering vol 1 no 5 2011
[6] Tutorial point with digital image processing httpwwwtutorialspointcomdiphistogram_equalizationhtm
[7] K Venkatalakshmi and S M Shalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and K-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Melbourne VIC Australia December2005
[8] K Venkatalakshmi and S M Shalinie ldquoMultispectral imageclassification using modified k-means clusteringrdquo In-ternational Journal on Neural and Mass-Parallel Computingand Information Systems vol 17 no 2 pp 113ndash120 2007
[9] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercy Shalinie ldquoMultispectral image clustering using en-hanced genetic k-means algorithmrdquo Information TechnologyJournal vol 6 no 4 pp 554ndash560 2007
[10] P I Dalatu ldquoTime complexity of k-means and k-mediansclustering algorithms in outliers detectionrdquo Global Journal ofPure and Applied Mathematics vol 12 no 5 pp 4405ndash44182016
[11] K Senthilkumar K Venkatalakshmi K Karthikeyan andP Kathirkamasundari ldquoAn efficient method for segmentingdigital image using a hybrid model of particle swarm opti-mization and artificial bee colony algorithmrdquo InternationalJournal of Applied Engineering Research vol 10 pp 444ndash4492015
[12] K Venkatalakshmi and S Mercyshalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and k-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Bangalore India December 2005
[13] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercyshalinie ldquoA customized particle swarm optimizationfor classification of multispectral imagery based on featurefusionrdquo International Arab Journal of Information Technologyvol 5 no 4 pp 327ndash333 2008
[14] J C Bansal and P K Singh ldquoInertia weight strategies inparticle swarm optimizationrdquo in Proceedings of IEEE WorldCongress on Nature and Biologically Inspired Computing pp640ndash647 Salamanca Spain October 2011
[15] P K Patel V Sharma and K Gupta ldquoGuaranteed conver-gence particle swarm optimization using Personal Bestrdquo In-ternational Journal of Computer Applications vol 73 no 7pp 6ndash10 2013
[16] X Wang L Ge and X Li ldquoEvaluation of filters for Envi-satAsar speckle suppression in pasture areardquo ISPRS Annals ofthe Photogrammetry Remote Sensing and Spatial InformationSciences vol 1ndash7 pp 341ndash346 2012
[17] H Nagaveena D Devaraj and S C Prasanna KumarldquoVessels segmentation in diabetic retinopathy by adaptivemedian thresholdingrdquo International Journal of Science andTechnology vol 1 no 1 pp 17ndash22 2013
[18] A Asuntha N Singh and A Srinivasan ldquoPSO genetic op-timization and SVM algorithm used for lung cancer
detectionrdquo Journal of Chemical and Pharmaceutical Researchvol 8 no 6 pp 351ndash359 2016
[19] P Bhuvaneswari and A Brintha+erese ldquoDetection of cancerin lung with K-NN classification using genetic algorithmrdquoInternational Conference on Nanomaterials and Technologiesvol 10 pp 433ndash440 2014
16 Computational and Mathematical Methods in Medicine
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom
+e neighborhood mean value is measured for each andevery pixel and is replaced by the corresponding meanvalue +is process is repeated for every pixel in the image[5] All the pixels in the digital image are modified bysliding the operator over the entire range of pixels +esteps followed for the average filter are given inAlgorithm 2
23 Histogram Equalization Image enhancement is thetechnique which is used to improve the image quality Forbetter understanding and analysis it is mandatory to en-hance the contrast of medical images +e conventionalmethod used for this operation is histogram equalization Aminor adjustment on the intensity of image pixels is donein this method Each pixel is mapped to intensity pro-portional to its rank in the surrounding pixels +e stepsfollowed for histogram equalization are given in Algo-rithm 3 [6]
24 k-Means Clustering Algorithm +e simplest and con-ventional method in cluster analysis is the k-means clus-tering algorithm+is algorithm segregates the given datasetinto two or more clusters [7] +e accuracy of this methodcompletely depends on the selection of the cluster center Itis mandatory to select the optimum cluster center to get abetter result +e Euclidean distance is the general measureto segregate the dataset [8] Pixels are assigned to an indi-vidual cluster based on the Euclidean distance +e objectivefunction used in this algorithm is
J(v) 1113944C
i11113944
Ci
j1xi minus vj
1113874 11138752 (1)
where xi are the pixels vj are the cluster centers xi minus vj isthe Euclidean distance between xi and vjCi is the number ofdata points for the ith cluster and C is the number of clustercenters [9] +e steps followed for k-means clustering aregiven in Algorithm 4
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 13 Resultant images by k-median clustering
Computational and Mathematical Methods in Medicine 11
25 k-Median Clustering Algorithm +is is also a clusteringalgorithm slightly modified from the k-means algorithm Incentroid calculation instead of calculating the mean valuethe median value is considered +is algorithm significantlyreduces the error since there is no squared operation as inthe calculation of the Euclidean distance +e clustersformed by this method are more compact As an alternatethis approach uses the Lloyd-style iteration +e steps fol-lowed for k-median clustering are given in Algorithm 5 [10]
26 Particle Swarm Optimization PSO is a metaheuristicalgorithm used efficiently in medical image analysis [11] Itmimics the social behavior of the birds searching for food [12]+e fundamental idea of PSO is sharing and communicatingthe information In this approach each particle has initialposition and velocity Based on the fitness value the velocity
and position are updated +e relevant two equations in PSOto update the position and velocity are as follows [11 12]
v(t + 1) v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minus x(t)]
x(t + 1) x(t) + v(t + 1)
(2)
where r1 and r2 are the random numbers and the accel-eration coefficients c1 and c2 are two positive constants+e success of PSO relies on the fitness function +efollowing fitness function has been used for the presentstudy
maximizef 1113944n
i1
intercluster distanceintracluster distance
(3)
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 14 Resultant images by the PSO algorithm
12 Computational and Mathematical Methods in Medicine
where n is the number of clusters +e steps followed for theparticle swarm optimization are shown in Algorithm 6
27 Inertia-Weighted Particle Swarm Optimization +eexploration and exploitation in PSO are based on the inertiaweight +e basic PSO presented by Eberhart and Kennedyin 1995 has no inertia weight In 1998 Shi and Eberhartintroduced the concept of inertia weight by adding constantinertia weight +ey stated that a significant inertia weightfacilitates a global search while a small inertia weight fa-cilitates a local search [14] +is enhances the convergencerate and reduces the number of iterations Inertia weight lessthan 1 in general improves the results +e used methodimproves the convergence rate and saves the time taken andsome iterations
+e resulting velocity update equation becomes
v(t + 1) wlowast v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minusx(t)](4)
where w is the inertia weight with constant inertia weight w
07 and random inertia weight w 05 + rand()2
28 Guaranteed Convergence Particle Swarm Optimization+e GCPSO focuses on a new particle which deals with thecurrent best position in the region In this task this particleis treated as a member of the swarm and the velocity updateequation for this new particle is given as follows [15]
vφ(t + 1) xφ(t) + pbest(t) + ωvφ(t) + ρ(t)(1minus 2r) (5)
+e search ability is increased by the social part +is willimprove the random search in the area around the gbestposition +e random vector and diameter of the search area
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 15 Resultant images by IWPSO algorithm clustering
Computational and Mathematical Methods in Medicine 13
are r and ρ(t) respectively +e range of the random vectorlies between 0 and 1 +e diameter of the search area can beupdated using the following equation
ρ(t + 1)
2ρ(t) successesgt sc
115
1113874 1113875ρ(t) failuresgt fc
ρ(t) otherwise
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎩
(6)
where the terms successes and failures are defined asthe number of consecutive successes and failures re-spectively +e threshold parameters sc and fc are de-termined empirically Since it is hard to obtain a bettervalue in only a few iterations in a high-dimensional searchspace the recommended values are thus sc 15 and fc 5
On some benchmark tests the GCPSO has shown anexcellent performance of locating the minimal of a spaceafter unimodal with only a small amount of particles +esteps to be followed for the GCPSO are shown inAlgorithm 7
3 Performance Measures
Certain performance measures are used to evaluate theresults obtained from medical image segmentation +e listof performance measures used to assess the filter operation isshown in Figure 2 [16] Let If be the image after noise re-duction and I0 be the noisy image
Performance measures used for the evaluation of theresults of the segmentation algorithm are given in Figure 3[17]
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 16 Resultant images by GCPSO algorithm clustering
14 Computational and Mathematical Methods in Medicine
4 Results and Discussion
+e used methods are practically implemented usingMATLAB coding and the results were verified
In the preprocessing stage a comparison was donebetween the performance of median adaptive median andmean filters +e SSI and SMPI values are shown in Table 1and Figures 4 and 5 From the results it is evident that theadaptive median filter has accurate characteristics than themean and median filters for medical image segmentation
+e segmentation accuracy was measured using the truepositive rate true negative rate false positive rate and falsenegative rate by comparing the results from the algorithmwithmanual segmentation results +e practical results of the k-means clustering segmentation algorithm are shown in Table 2
+e practical results of the k-median clustering seg-mentation algorithm are shown in Table 3
+e practical results of the PSO-based segmentationalgorithm are shown in Table 4
+e practical results of the IWPSO segmentation algo-rithm are shown in Table 5
+e practical results of the GCPSO segmentation algo-rithm are shown in Table 6
+e graphical view of the comparison of the true positiverate true negative rate false positive rate and false negativerate for the algorithms used is shown in Figures 6ndash9 It isproved that the true positive and true negative rates are highand false positive and false negative rates are low for theGCPSO algorithm
+e comparative evaluation based on the accuracy of thesegmentation is shown in Table 7 and Figure 10 +e resultsindicate that the GCPSO-based technique has the highestaverage value of accuracy than the other methods
+e resultant images after preprocessing are shown inFigures 11(a) and 11(b)
+e resultant images after segmentation using k-meansclustering are shown in Figure 12
+e resultant images after segmentation using k-medianclustering are shown in Figure 13
+e resultant images after segmentation using the PSOalgorithm are shown in Figure 14
+e resultant images after segmentation using theIWPSO algorithm are shown in Figure 15
+e resultant images after segmentation using theGCPSO algorithm are shown in Figure 16
In an earlier research lung cancer detection was doneusing PSO genetic optimization and SVM algorithm withthe Gabor filter and produced an accuracy of 895 [18]+emethod to detect lung cancer by means of K-NN classifi-cation using the genetic algorithm produced a maximumaccuracy of 90 [19]+e comparative results with respect tothe above-said methods are shown in Table 8
+e graphical comparative analysis between the used andexisting methods is shown in Figure 17
5 Conclusion
In this study various optimization algorithms have beenevaluated to detect the tumor Medical images often need
preprocessing before being subjected to statistical analysis+e adaptive median filter has better results than medianand mean filters because the speckle suppression index andspeckle and mean preservation index values are lower for theadaptive median filter Comparing the five algorithms theaccuracy of the tumor extraction is improved in GCPSOwith the highest accuracy of 958079 and it obtained above90 of precision in all the 20 images It is more accuratewhen compared to the previous method which had an ac-curacy of 90 in 4 out of 10 datasets only In future studiesthe use of more number of optimization algorithms will beincluded to improve the accuracy
Data Availability
+eCTimages data used to support the findings of this studyhave been deposited in the LungCT-Diagnosis repository(doiorg107937K9TCIA2015A6V7JIWX)
Conflicts of Interest
+e authors declare that there are no conflicts of interestregarding the publication of this paper
References
[1] A A Brindha S Indirani and A Srinivasan ldquoLung cancerdetection using SVM algorithm and optimization techniquesrdquoJournal of Chemical and Pharmaceutical Sciences vol 9 no 42016
[2] M Kurkure and A +akare ldquoIntroducing automated systemfor lung cancer detection using Evolutionary ApproachrdquoInternational Journal of Engineering and Computer Sciencevol 5 no 5 pp 16736ndash16739 2016
[3] B Rani A K Goel and R Kaur ldquoA modified approach forlung cancer detection using bacterial forging optimization
868890929496
PSO GA and SVM algorithm
K-NNclassification
using GA
Proposed GCPSOmethod
8950 90
9581
Acc
urac
y
Various algorithms
Comparison between existing and projected methods
Figure 17 Graphical view of accuracy
Table 8 Comparative analysis of accuracy of the projected methodwith various methods
Various methods Accuracy ()PSO GA and SVM algorithm [18] 8950K-NN classification using GA [19] 9000Projected GCPSO method 9581
Computational and Mathematical Methods in Medicine 15
algorithmrdquo International Journal of Scientific Research En-gineering and Technology vol 5 no 1 2016
[4] N Panpaliya N Tadas S Bobade R Aglawe andA Gudadhe ldquoA survey on early detection and prediction oflung cancerrdquo International Journal of Computer Science andMobile Computing vol 4 no 1 pp 175ndash184 2015
[5] G Gupta ldquoAlgorithm for image processing using improvedmedian filter and comparison of mean median and improvedmedian filterrdquo International Journal of Soft Computing andEngineering vol 1 no 5 2011
[6] Tutorial point with digital image processing httpwwwtutorialspointcomdiphistogram_equalizationhtm
[7] K Venkatalakshmi and S M Shalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and K-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Melbourne VIC Australia December2005
[8] K Venkatalakshmi and S M Shalinie ldquoMultispectral imageclassification using modified k-means clusteringrdquo In-ternational Journal on Neural and Mass-Parallel Computingand Information Systems vol 17 no 2 pp 113ndash120 2007
[9] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercy Shalinie ldquoMultispectral image clustering using en-hanced genetic k-means algorithmrdquo Information TechnologyJournal vol 6 no 4 pp 554ndash560 2007
[10] P I Dalatu ldquoTime complexity of k-means and k-mediansclustering algorithms in outliers detectionrdquo Global Journal ofPure and Applied Mathematics vol 12 no 5 pp 4405ndash44182016
[11] K Senthilkumar K Venkatalakshmi K Karthikeyan andP Kathirkamasundari ldquoAn efficient method for segmentingdigital image using a hybrid model of particle swarm opti-mization and artificial bee colony algorithmrdquo InternationalJournal of Applied Engineering Research vol 10 pp 444ndash4492015
[12] K Venkatalakshmi and S Mercyshalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and k-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Bangalore India December 2005
[13] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercyshalinie ldquoA customized particle swarm optimizationfor classification of multispectral imagery based on featurefusionrdquo International Arab Journal of Information Technologyvol 5 no 4 pp 327ndash333 2008
[14] J C Bansal and P K Singh ldquoInertia weight strategies inparticle swarm optimizationrdquo in Proceedings of IEEE WorldCongress on Nature and Biologically Inspired Computing pp640ndash647 Salamanca Spain October 2011
[15] P K Patel V Sharma and K Gupta ldquoGuaranteed conver-gence particle swarm optimization using Personal Bestrdquo In-ternational Journal of Computer Applications vol 73 no 7pp 6ndash10 2013
[16] X Wang L Ge and X Li ldquoEvaluation of filters for Envi-satAsar speckle suppression in pasture areardquo ISPRS Annals ofthe Photogrammetry Remote Sensing and Spatial InformationSciences vol 1ndash7 pp 341ndash346 2012
[17] H Nagaveena D Devaraj and S C Prasanna KumarldquoVessels segmentation in diabetic retinopathy by adaptivemedian thresholdingrdquo International Journal of Science andTechnology vol 1 no 1 pp 17ndash22 2013
[18] A Asuntha N Singh and A Srinivasan ldquoPSO genetic op-timization and SVM algorithm used for lung cancer
detectionrdquo Journal of Chemical and Pharmaceutical Researchvol 8 no 6 pp 351ndash359 2016
[19] P Bhuvaneswari and A Brintha+erese ldquoDetection of cancerin lung with K-NN classification using genetic algorithmrdquoInternational Conference on Nanomaterials and Technologiesvol 10 pp 433ndash440 2014
16 Computational and Mathematical Methods in Medicine
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom
25 k-Median Clustering Algorithm +is is also a clusteringalgorithm slightly modified from the k-means algorithm Incentroid calculation instead of calculating the mean valuethe median value is considered +is algorithm significantlyreduces the error since there is no squared operation as inthe calculation of the Euclidean distance +e clustersformed by this method are more compact As an alternatethis approach uses the Lloyd-style iteration +e steps fol-lowed for k-median clustering are given in Algorithm 5 [10]
26 Particle Swarm Optimization PSO is a metaheuristicalgorithm used efficiently in medical image analysis [11] Itmimics the social behavior of the birds searching for food [12]+e fundamental idea of PSO is sharing and communicatingthe information In this approach each particle has initialposition and velocity Based on the fitness value the velocity
and position are updated +e relevant two equations in PSOto update the position and velocity are as follows [11 12]
v(t + 1) v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minus x(t)]
x(t + 1) x(t) + v(t + 1)
(2)
where r1 and r2 are the random numbers and the accel-eration coefficients c1 and c2 are two positive constants+e success of PSO relies on the fitness function +efollowing fitness function has been used for the presentstudy
maximizef 1113944n
i1
intercluster distanceintracluster distance
(3)
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 14 Resultant images by the PSO algorithm
12 Computational and Mathematical Methods in Medicine
where n is the number of clusters +e steps followed for theparticle swarm optimization are shown in Algorithm 6
27 Inertia-Weighted Particle Swarm Optimization +eexploration and exploitation in PSO are based on the inertiaweight +e basic PSO presented by Eberhart and Kennedyin 1995 has no inertia weight In 1998 Shi and Eberhartintroduced the concept of inertia weight by adding constantinertia weight +ey stated that a significant inertia weightfacilitates a global search while a small inertia weight fa-cilitates a local search [14] +is enhances the convergencerate and reduces the number of iterations Inertia weight lessthan 1 in general improves the results +e used methodimproves the convergence rate and saves the time taken andsome iterations
+e resulting velocity update equation becomes
v(t + 1) wlowast v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minusx(t)](4)
where w is the inertia weight with constant inertia weight w
07 and random inertia weight w 05 + rand()2
28 Guaranteed Convergence Particle Swarm Optimization+e GCPSO focuses on a new particle which deals with thecurrent best position in the region In this task this particleis treated as a member of the swarm and the velocity updateequation for this new particle is given as follows [15]
vφ(t + 1) xφ(t) + pbest(t) + ωvφ(t) + ρ(t)(1minus 2r) (5)
+e search ability is increased by the social part +is willimprove the random search in the area around the gbestposition +e random vector and diameter of the search area
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 15 Resultant images by IWPSO algorithm clustering
Computational and Mathematical Methods in Medicine 13
are r and ρ(t) respectively +e range of the random vectorlies between 0 and 1 +e diameter of the search area can beupdated using the following equation
ρ(t + 1)
2ρ(t) successesgt sc
115
1113874 1113875ρ(t) failuresgt fc
ρ(t) otherwise
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎩
(6)
where the terms successes and failures are defined asthe number of consecutive successes and failures re-spectively +e threshold parameters sc and fc are de-termined empirically Since it is hard to obtain a bettervalue in only a few iterations in a high-dimensional searchspace the recommended values are thus sc 15 and fc 5
On some benchmark tests the GCPSO has shown anexcellent performance of locating the minimal of a spaceafter unimodal with only a small amount of particles +esteps to be followed for the GCPSO are shown inAlgorithm 7
3 Performance Measures
Certain performance measures are used to evaluate theresults obtained from medical image segmentation +e listof performance measures used to assess the filter operation isshown in Figure 2 [16] Let If be the image after noise re-duction and I0 be the noisy image
Performance measures used for the evaluation of theresults of the segmentation algorithm are given in Figure 3[17]
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 16 Resultant images by GCPSO algorithm clustering
14 Computational and Mathematical Methods in Medicine
4 Results and Discussion
+e used methods are practically implemented usingMATLAB coding and the results were verified
In the preprocessing stage a comparison was donebetween the performance of median adaptive median andmean filters +e SSI and SMPI values are shown in Table 1and Figures 4 and 5 From the results it is evident that theadaptive median filter has accurate characteristics than themean and median filters for medical image segmentation
+e segmentation accuracy was measured using the truepositive rate true negative rate false positive rate and falsenegative rate by comparing the results from the algorithmwithmanual segmentation results +e practical results of the k-means clustering segmentation algorithm are shown in Table 2
+e practical results of the k-median clustering seg-mentation algorithm are shown in Table 3
+e practical results of the PSO-based segmentationalgorithm are shown in Table 4
+e practical results of the IWPSO segmentation algo-rithm are shown in Table 5
+e practical results of the GCPSO segmentation algo-rithm are shown in Table 6
+e graphical view of the comparison of the true positiverate true negative rate false positive rate and false negativerate for the algorithms used is shown in Figures 6ndash9 It isproved that the true positive and true negative rates are highand false positive and false negative rates are low for theGCPSO algorithm
+e comparative evaluation based on the accuracy of thesegmentation is shown in Table 7 and Figure 10 +e resultsindicate that the GCPSO-based technique has the highestaverage value of accuracy than the other methods
+e resultant images after preprocessing are shown inFigures 11(a) and 11(b)
+e resultant images after segmentation using k-meansclustering are shown in Figure 12
+e resultant images after segmentation using k-medianclustering are shown in Figure 13
+e resultant images after segmentation using the PSOalgorithm are shown in Figure 14
+e resultant images after segmentation using theIWPSO algorithm are shown in Figure 15
+e resultant images after segmentation using theGCPSO algorithm are shown in Figure 16
In an earlier research lung cancer detection was doneusing PSO genetic optimization and SVM algorithm withthe Gabor filter and produced an accuracy of 895 [18]+emethod to detect lung cancer by means of K-NN classifi-cation using the genetic algorithm produced a maximumaccuracy of 90 [19]+e comparative results with respect tothe above-said methods are shown in Table 8
+e graphical comparative analysis between the used andexisting methods is shown in Figure 17
5 Conclusion
In this study various optimization algorithms have beenevaluated to detect the tumor Medical images often need
preprocessing before being subjected to statistical analysis+e adaptive median filter has better results than medianand mean filters because the speckle suppression index andspeckle and mean preservation index values are lower for theadaptive median filter Comparing the five algorithms theaccuracy of the tumor extraction is improved in GCPSOwith the highest accuracy of 958079 and it obtained above90 of precision in all the 20 images It is more accuratewhen compared to the previous method which had an ac-curacy of 90 in 4 out of 10 datasets only In future studiesthe use of more number of optimization algorithms will beincluded to improve the accuracy
Data Availability
+eCTimages data used to support the findings of this studyhave been deposited in the LungCT-Diagnosis repository(doiorg107937K9TCIA2015A6V7JIWX)
Conflicts of Interest
+e authors declare that there are no conflicts of interestregarding the publication of this paper
References
[1] A A Brindha S Indirani and A Srinivasan ldquoLung cancerdetection using SVM algorithm and optimization techniquesrdquoJournal of Chemical and Pharmaceutical Sciences vol 9 no 42016
[2] M Kurkure and A +akare ldquoIntroducing automated systemfor lung cancer detection using Evolutionary ApproachrdquoInternational Journal of Engineering and Computer Sciencevol 5 no 5 pp 16736ndash16739 2016
[3] B Rani A K Goel and R Kaur ldquoA modified approach forlung cancer detection using bacterial forging optimization
868890929496
PSO GA and SVM algorithm
K-NNclassification
using GA
Proposed GCPSOmethod
8950 90
9581
Acc
urac
y
Various algorithms
Comparison between existing and projected methods
Figure 17 Graphical view of accuracy
Table 8 Comparative analysis of accuracy of the projected methodwith various methods
Various methods Accuracy ()PSO GA and SVM algorithm [18] 8950K-NN classification using GA [19] 9000Projected GCPSO method 9581
Computational and Mathematical Methods in Medicine 15
algorithmrdquo International Journal of Scientific Research En-gineering and Technology vol 5 no 1 2016
[4] N Panpaliya N Tadas S Bobade R Aglawe andA Gudadhe ldquoA survey on early detection and prediction oflung cancerrdquo International Journal of Computer Science andMobile Computing vol 4 no 1 pp 175ndash184 2015
[5] G Gupta ldquoAlgorithm for image processing using improvedmedian filter and comparison of mean median and improvedmedian filterrdquo International Journal of Soft Computing andEngineering vol 1 no 5 2011
[6] Tutorial point with digital image processing httpwwwtutorialspointcomdiphistogram_equalizationhtm
[7] K Venkatalakshmi and S M Shalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and K-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Melbourne VIC Australia December2005
[8] K Venkatalakshmi and S M Shalinie ldquoMultispectral imageclassification using modified k-means clusteringrdquo In-ternational Journal on Neural and Mass-Parallel Computingand Information Systems vol 17 no 2 pp 113ndash120 2007
[9] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercy Shalinie ldquoMultispectral image clustering using en-hanced genetic k-means algorithmrdquo Information TechnologyJournal vol 6 no 4 pp 554ndash560 2007
[10] P I Dalatu ldquoTime complexity of k-means and k-mediansclustering algorithms in outliers detectionrdquo Global Journal ofPure and Applied Mathematics vol 12 no 5 pp 4405ndash44182016
[11] K Senthilkumar K Venkatalakshmi K Karthikeyan andP Kathirkamasundari ldquoAn efficient method for segmentingdigital image using a hybrid model of particle swarm opti-mization and artificial bee colony algorithmrdquo InternationalJournal of Applied Engineering Research vol 10 pp 444ndash4492015
[12] K Venkatalakshmi and S Mercyshalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and k-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Bangalore India December 2005
[13] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercyshalinie ldquoA customized particle swarm optimizationfor classification of multispectral imagery based on featurefusionrdquo International Arab Journal of Information Technologyvol 5 no 4 pp 327ndash333 2008
[14] J C Bansal and P K Singh ldquoInertia weight strategies inparticle swarm optimizationrdquo in Proceedings of IEEE WorldCongress on Nature and Biologically Inspired Computing pp640ndash647 Salamanca Spain October 2011
[15] P K Patel V Sharma and K Gupta ldquoGuaranteed conver-gence particle swarm optimization using Personal Bestrdquo In-ternational Journal of Computer Applications vol 73 no 7pp 6ndash10 2013
[16] X Wang L Ge and X Li ldquoEvaluation of filters for Envi-satAsar speckle suppression in pasture areardquo ISPRS Annals ofthe Photogrammetry Remote Sensing and Spatial InformationSciences vol 1ndash7 pp 341ndash346 2012
[17] H Nagaveena D Devaraj and S C Prasanna KumarldquoVessels segmentation in diabetic retinopathy by adaptivemedian thresholdingrdquo International Journal of Science andTechnology vol 1 no 1 pp 17ndash22 2013
[18] A Asuntha N Singh and A Srinivasan ldquoPSO genetic op-timization and SVM algorithm used for lung cancer
detectionrdquo Journal of Chemical and Pharmaceutical Researchvol 8 no 6 pp 351ndash359 2016
[19] P Bhuvaneswari and A Brintha+erese ldquoDetection of cancerin lung with K-NN classification using genetic algorithmrdquoInternational Conference on Nanomaterials and Technologiesvol 10 pp 433ndash440 2014
16 Computational and Mathematical Methods in Medicine
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom
where n is the number of clusters +e steps followed for theparticle swarm optimization are shown in Algorithm 6
27 Inertia-Weighted Particle Swarm Optimization +eexploration and exploitation in PSO are based on the inertiaweight +e basic PSO presented by Eberhart and Kennedyin 1995 has no inertia weight In 1998 Shi and Eberhartintroduced the concept of inertia weight by adding constantinertia weight +ey stated that a significant inertia weightfacilitates a global search while a small inertia weight fa-cilitates a local search [14] +is enhances the convergencerate and reduces the number of iterations Inertia weight lessthan 1 in general improves the results +e used methodimproves the convergence rate and saves the time taken andsome iterations
+e resulting velocity update equation becomes
v(t + 1) wlowast v(t) + c1r1[pbest(t)minusx(t)]
+ c2r2[gbest(t)minusx(t)](4)
where w is the inertia weight with constant inertia weight w
07 and random inertia weight w 05 + rand()2
28 Guaranteed Convergence Particle Swarm Optimization+e GCPSO focuses on a new particle which deals with thecurrent best position in the region In this task this particleis treated as a member of the swarm and the velocity updateequation for this new particle is given as follows [15]
vφ(t + 1) xφ(t) + pbest(t) + ωvφ(t) + ρ(t)(1minus 2r) (5)
+e search ability is increased by the social part +is willimprove the random search in the area around the gbestposition +e random vector and diameter of the search area
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 15 Resultant images by IWPSO algorithm clustering
Computational and Mathematical Methods in Medicine 13
are r and ρ(t) respectively +e range of the random vectorlies between 0 and 1 +e diameter of the search area can beupdated using the following equation
ρ(t + 1)
2ρ(t) successesgt sc
115
1113874 1113875ρ(t) failuresgt fc
ρ(t) otherwise
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎩
(6)
where the terms successes and failures are defined asthe number of consecutive successes and failures re-spectively +e threshold parameters sc and fc are de-termined empirically Since it is hard to obtain a bettervalue in only a few iterations in a high-dimensional searchspace the recommended values are thus sc 15 and fc 5
On some benchmark tests the GCPSO has shown anexcellent performance of locating the minimal of a spaceafter unimodal with only a small amount of particles +esteps to be followed for the GCPSO are shown inAlgorithm 7
3 Performance Measures
Certain performance measures are used to evaluate theresults obtained from medical image segmentation +e listof performance measures used to assess the filter operation isshown in Figure 2 [16] Let If be the image after noise re-duction and I0 be the noisy image
Performance measures used for the evaluation of theresults of the segmentation algorithm are given in Figure 3[17]
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 16 Resultant images by GCPSO algorithm clustering
14 Computational and Mathematical Methods in Medicine
4 Results and Discussion
+e used methods are practically implemented usingMATLAB coding and the results were verified
In the preprocessing stage a comparison was donebetween the performance of median adaptive median andmean filters +e SSI and SMPI values are shown in Table 1and Figures 4 and 5 From the results it is evident that theadaptive median filter has accurate characteristics than themean and median filters for medical image segmentation
+e segmentation accuracy was measured using the truepositive rate true negative rate false positive rate and falsenegative rate by comparing the results from the algorithmwithmanual segmentation results +e practical results of the k-means clustering segmentation algorithm are shown in Table 2
+e practical results of the k-median clustering seg-mentation algorithm are shown in Table 3
+e practical results of the PSO-based segmentationalgorithm are shown in Table 4
+e practical results of the IWPSO segmentation algo-rithm are shown in Table 5
+e practical results of the GCPSO segmentation algo-rithm are shown in Table 6
+e graphical view of the comparison of the true positiverate true negative rate false positive rate and false negativerate for the algorithms used is shown in Figures 6ndash9 It isproved that the true positive and true negative rates are highand false positive and false negative rates are low for theGCPSO algorithm
+e comparative evaluation based on the accuracy of thesegmentation is shown in Table 7 and Figure 10 +e resultsindicate that the GCPSO-based technique has the highestaverage value of accuracy than the other methods
+e resultant images after preprocessing are shown inFigures 11(a) and 11(b)
+e resultant images after segmentation using k-meansclustering are shown in Figure 12
+e resultant images after segmentation using k-medianclustering are shown in Figure 13
+e resultant images after segmentation using the PSOalgorithm are shown in Figure 14
+e resultant images after segmentation using theIWPSO algorithm are shown in Figure 15
+e resultant images after segmentation using theGCPSO algorithm are shown in Figure 16
In an earlier research lung cancer detection was doneusing PSO genetic optimization and SVM algorithm withthe Gabor filter and produced an accuracy of 895 [18]+emethod to detect lung cancer by means of K-NN classifi-cation using the genetic algorithm produced a maximumaccuracy of 90 [19]+e comparative results with respect tothe above-said methods are shown in Table 8
+e graphical comparative analysis between the used andexisting methods is shown in Figure 17
5 Conclusion
In this study various optimization algorithms have beenevaluated to detect the tumor Medical images often need
preprocessing before being subjected to statistical analysis+e adaptive median filter has better results than medianand mean filters because the speckle suppression index andspeckle and mean preservation index values are lower for theadaptive median filter Comparing the five algorithms theaccuracy of the tumor extraction is improved in GCPSOwith the highest accuracy of 958079 and it obtained above90 of precision in all the 20 images It is more accuratewhen compared to the previous method which had an ac-curacy of 90 in 4 out of 10 datasets only In future studiesthe use of more number of optimization algorithms will beincluded to improve the accuracy
Data Availability
+eCTimages data used to support the findings of this studyhave been deposited in the LungCT-Diagnosis repository(doiorg107937K9TCIA2015A6V7JIWX)
Conflicts of Interest
+e authors declare that there are no conflicts of interestregarding the publication of this paper
References
[1] A A Brindha S Indirani and A Srinivasan ldquoLung cancerdetection using SVM algorithm and optimization techniquesrdquoJournal of Chemical and Pharmaceutical Sciences vol 9 no 42016
[2] M Kurkure and A +akare ldquoIntroducing automated systemfor lung cancer detection using Evolutionary ApproachrdquoInternational Journal of Engineering and Computer Sciencevol 5 no 5 pp 16736ndash16739 2016
[3] B Rani A K Goel and R Kaur ldquoA modified approach forlung cancer detection using bacterial forging optimization
868890929496
PSO GA and SVM algorithm
K-NNclassification
using GA
Proposed GCPSOmethod
8950 90
9581
Acc
urac
y
Various algorithms
Comparison between existing and projected methods
Figure 17 Graphical view of accuracy
Table 8 Comparative analysis of accuracy of the projected methodwith various methods
Various methods Accuracy ()PSO GA and SVM algorithm [18] 8950K-NN classification using GA [19] 9000Projected GCPSO method 9581
Computational and Mathematical Methods in Medicine 15
algorithmrdquo International Journal of Scientific Research En-gineering and Technology vol 5 no 1 2016
[4] N Panpaliya N Tadas S Bobade R Aglawe andA Gudadhe ldquoA survey on early detection and prediction oflung cancerrdquo International Journal of Computer Science andMobile Computing vol 4 no 1 pp 175ndash184 2015
[5] G Gupta ldquoAlgorithm for image processing using improvedmedian filter and comparison of mean median and improvedmedian filterrdquo International Journal of Soft Computing andEngineering vol 1 no 5 2011
[6] Tutorial point with digital image processing httpwwwtutorialspointcomdiphistogram_equalizationhtm
[7] K Venkatalakshmi and S M Shalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and K-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Melbourne VIC Australia December2005
[8] K Venkatalakshmi and S M Shalinie ldquoMultispectral imageclassification using modified k-means clusteringrdquo In-ternational Journal on Neural and Mass-Parallel Computingand Information Systems vol 17 no 2 pp 113ndash120 2007
[9] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercy Shalinie ldquoMultispectral image clustering using en-hanced genetic k-means algorithmrdquo Information TechnologyJournal vol 6 no 4 pp 554ndash560 2007
[10] P I Dalatu ldquoTime complexity of k-means and k-mediansclustering algorithms in outliers detectionrdquo Global Journal ofPure and Applied Mathematics vol 12 no 5 pp 4405ndash44182016
[11] K Senthilkumar K Venkatalakshmi K Karthikeyan andP Kathirkamasundari ldquoAn efficient method for segmentingdigital image using a hybrid model of particle swarm opti-mization and artificial bee colony algorithmrdquo InternationalJournal of Applied Engineering Research vol 10 pp 444ndash4492015
[12] K Venkatalakshmi and S Mercyshalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and k-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Bangalore India December 2005
[13] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercyshalinie ldquoA customized particle swarm optimizationfor classification of multispectral imagery based on featurefusionrdquo International Arab Journal of Information Technologyvol 5 no 4 pp 327ndash333 2008
[14] J C Bansal and P K Singh ldquoInertia weight strategies inparticle swarm optimizationrdquo in Proceedings of IEEE WorldCongress on Nature and Biologically Inspired Computing pp640ndash647 Salamanca Spain October 2011
[15] P K Patel V Sharma and K Gupta ldquoGuaranteed conver-gence particle swarm optimization using Personal Bestrdquo In-ternational Journal of Computer Applications vol 73 no 7pp 6ndash10 2013
[16] X Wang L Ge and X Li ldquoEvaluation of filters for Envi-satAsar speckle suppression in pasture areardquo ISPRS Annals ofthe Photogrammetry Remote Sensing and Spatial InformationSciences vol 1ndash7 pp 341ndash346 2012
[17] H Nagaveena D Devaraj and S C Prasanna KumarldquoVessels segmentation in diabetic retinopathy by adaptivemedian thresholdingrdquo International Journal of Science andTechnology vol 1 no 1 pp 17ndash22 2013
[18] A Asuntha N Singh and A Srinivasan ldquoPSO genetic op-timization and SVM algorithm used for lung cancer
detectionrdquo Journal of Chemical and Pharmaceutical Researchvol 8 no 6 pp 351ndash359 2016
[19] P Bhuvaneswari and A Brintha+erese ldquoDetection of cancerin lung with K-NN classification using genetic algorithmrdquoInternational Conference on Nanomaterials and Technologiesvol 10 pp 433ndash440 2014
16 Computational and Mathematical Methods in Medicine
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom
are r and ρ(t) respectively +e range of the random vectorlies between 0 and 1 +e diameter of the search area can beupdated using the following equation
ρ(t + 1)
2ρ(t) successesgt sc
115
1113874 1113875ρ(t) failuresgt fc
ρ(t) otherwise
⎧⎪⎪⎪⎪⎪⎪⎪⎪⎨
⎪⎪⎪⎪⎪⎪⎪⎪⎩
(6)
where the terms successes and failures are defined asthe number of consecutive successes and failures re-spectively +e threshold parameters sc and fc are de-termined empirically Since it is hard to obtain a bettervalue in only a few iterations in a high-dimensional searchspace the recommended values are thus sc 15 and fc 5
On some benchmark tests the GCPSO has shown anexcellent performance of locating the minimal of a spaceafter unimodal with only a small amount of particles +esteps to be followed for the GCPSO are shown inAlgorithm 7
3 Performance Measures
Certain performance measures are used to evaluate theresults obtained from medical image segmentation +e listof performance measures used to assess the filter operation isshown in Figure 2 [16] Let If be the image after noise re-duction and I0 be the noisy image
Performance measures used for the evaluation of theresults of the segmentation algorithm are given in Figure 3[17]
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Image 01
Image 02
Image 03
Image 04
Image 05
Image 06
Image 07
Image 08
Image 09
Image 10
Clusteredimage
Segmentedimage
Extractedtumor
Manualextraction
Sampleimages
Figure 16 Resultant images by GCPSO algorithm clustering
14 Computational and Mathematical Methods in Medicine
4 Results and Discussion
+e used methods are practically implemented usingMATLAB coding and the results were verified
In the preprocessing stage a comparison was donebetween the performance of median adaptive median andmean filters +e SSI and SMPI values are shown in Table 1and Figures 4 and 5 From the results it is evident that theadaptive median filter has accurate characteristics than themean and median filters for medical image segmentation
+e segmentation accuracy was measured using the truepositive rate true negative rate false positive rate and falsenegative rate by comparing the results from the algorithmwithmanual segmentation results +e practical results of the k-means clustering segmentation algorithm are shown in Table 2
+e practical results of the k-median clustering seg-mentation algorithm are shown in Table 3
+e practical results of the PSO-based segmentationalgorithm are shown in Table 4
+e practical results of the IWPSO segmentation algo-rithm are shown in Table 5
+e practical results of the GCPSO segmentation algo-rithm are shown in Table 6
+e graphical view of the comparison of the true positiverate true negative rate false positive rate and false negativerate for the algorithms used is shown in Figures 6ndash9 It isproved that the true positive and true negative rates are highand false positive and false negative rates are low for theGCPSO algorithm
+e comparative evaluation based on the accuracy of thesegmentation is shown in Table 7 and Figure 10 +e resultsindicate that the GCPSO-based technique has the highestaverage value of accuracy than the other methods
+e resultant images after preprocessing are shown inFigures 11(a) and 11(b)
+e resultant images after segmentation using k-meansclustering are shown in Figure 12
+e resultant images after segmentation using k-medianclustering are shown in Figure 13
+e resultant images after segmentation using the PSOalgorithm are shown in Figure 14
+e resultant images after segmentation using theIWPSO algorithm are shown in Figure 15
+e resultant images after segmentation using theGCPSO algorithm are shown in Figure 16
In an earlier research lung cancer detection was doneusing PSO genetic optimization and SVM algorithm withthe Gabor filter and produced an accuracy of 895 [18]+emethod to detect lung cancer by means of K-NN classifi-cation using the genetic algorithm produced a maximumaccuracy of 90 [19]+e comparative results with respect tothe above-said methods are shown in Table 8
+e graphical comparative analysis between the used andexisting methods is shown in Figure 17
5 Conclusion
In this study various optimization algorithms have beenevaluated to detect the tumor Medical images often need
preprocessing before being subjected to statistical analysis+e adaptive median filter has better results than medianand mean filters because the speckle suppression index andspeckle and mean preservation index values are lower for theadaptive median filter Comparing the five algorithms theaccuracy of the tumor extraction is improved in GCPSOwith the highest accuracy of 958079 and it obtained above90 of precision in all the 20 images It is more accuratewhen compared to the previous method which had an ac-curacy of 90 in 4 out of 10 datasets only In future studiesthe use of more number of optimization algorithms will beincluded to improve the accuracy
Data Availability
+eCTimages data used to support the findings of this studyhave been deposited in the LungCT-Diagnosis repository(doiorg107937K9TCIA2015A6V7JIWX)
Conflicts of Interest
+e authors declare that there are no conflicts of interestregarding the publication of this paper
References
[1] A A Brindha S Indirani and A Srinivasan ldquoLung cancerdetection using SVM algorithm and optimization techniquesrdquoJournal of Chemical and Pharmaceutical Sciences vol 9 no 42016
[2] M Kurkure and A +akare ldquoIntroducing automated systemfor lung cancer detection using Evolutionary ApproachrdquoInternational Journal of Engineering and Computer Sciencevol 5 no 5 pp 16736ndash16739 2016
[3] B Rani A K Goel and R Kaur ldquoA modified approach forlung cancer detection using bacterial forging optimization
868890929496
PSO GA and SVM algorithm
K-NNclassification
using GA
Proposed GCPSOmethod
8950 90
9581
Acc
urac
y
Various algorithms
Comparison between existing and projected methods
Figure 17 Graphical view of accuracy
Table 8 Comparative analysis of accuracy of the projected methodwith various methods
Various methods Accuracy ()PSO GA and SVM algorithm [18] 8950K-NN classification using GA [19] 9000Projected GCPSO method 9581
Computational and Mathematical Methods in Medicine 15
algorithmrdquo International Journal of Scientific Research En-gineering and Technology vol 5 no 1 2016
[4] N Panpaliya N Tadas S Bobade R Aglawe andA Gudadhe ldquoA survey on early detection and prediction oflung cancerrdquo International Journal of Computer Science andMobile Computing vol 4 no 1 pp 175ndash184 2015
[5] G Gupta ldquoAlgorithm for image processing using improvedmedian filter and comparison of mean median and improvedmedian filterrdquo International Journal of Soft Computing andEngineering vol 1 no 5 2011
[6] Tutorial point with digital image processing httpwwwtutorialspointcomdiphistogram_equalizationhtm
[7] K Venkatalakshmi and S M Shalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and K-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Melbourne VIC Australia December2005
[8] K Venkatalakshmi and S M Shalinie ldquoMultispectral imageclassification using modified k-means clusteringrdquo In-ternational Journal on Neural and Mass-Parallel Computingand Information Systems vol 17 no 2 pp 113ndash120 2007
[9] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercy Shalinie ldquoMultispectral image clustering using en-hanced genetic k-means algorithmrdquo Information TechnologyJournal vol 6 no 4 pp 554ndash560 2007
[10] P I Dalatu ldquoTime complexity of k-means and k-mediansclustering algorithms in outliers detectionrdquo Global Journal ofPure and Applied Mathematics vol 12 no 5 pp 4405ndash44182016
[11] K Senthilkumar K Venkatalakshmi K Karthikeyan andP Kathirkamasundari ldquoAn efficient method for segmentingdigital image using a hybrid model of particle swarm opti-mization and artificial bee colony algorithmrdquo InternationalJournal of Applied Engineering Research vol 10 pp 444ndash4492015
[12] K Venkatalakshmi and S Mercyshalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and k-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Bangalore India December 2005
[13] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercyshalinie ldquoA customized particle swarm optimizationfor classification of multispectral imagery based on featurefusionrdquo International Arab Journal of Information Technologyvol 5 no 4 pp 327ndash333 2008
[14] J C Bansal and P K Singh ldquoInertia weight strategies inparticle swarm optimizationrdquo in Proceedings of IEEE WorldCongress on Nature and Biologically Inspired Computing pp640ndash647 Salamanca Spain October 2011
[15] P K Patel V Sharma and K Gupta ldquoGuaranteed conver-gence particle swarm optimization using Personal Bestrdquo In-ternational Journal of Computer Applications vol 73 no 7pp 6ndash10 2013
[16] X Wang L Ge and X Li ldquoEvaluation of filters for Envi-satAsar speckle suppression in pasture areardquo ISPRS Annals ofthe Photogrammetry Remote Sensing and Spatial InformationSciences vol 1ndash7 pp 341ndash346 2012
[17] H Nagaveena D Devaraj and S C Prasanna KumarldquoVessels segmentation in diabetic retinopathy by adaptivemedian thresholdingrdquo International Journal of Science andTechnology vol 1 no 1 pp 17ndash22 2013
[18] A Asuntha N Singh and A Srinivasan ldquoPSO genetic op-timization and SVM algorithm used for lung cancer
detectionrdquo Journal of Chemical and Pharmaceutical Researchvol 8 no 6 pp 351ndash359 2016
[19] P Bhuvaneswari and A Brintha+erese ldquoDetection of cancerin lung with K-NN classification using genetic algorithmrdquoInternational Conference on Nanomaterials and Technologiesvol 10 pp 433ndash440 2014
16 Computational and Mathematical Methods in Medicine
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom
4 Results and Discussion
+e used methods are practically implemented usingMATLAB coding and the results were verified
In the preprocessing stage a comparison was donebetween the performance of median adaptive median andmean filters +e SSI and SMPI values are shown in Table 1and Figures 4 and 5 From the results it is evident that theadaptive median filter has accurate characteristics than themean and median filters for medical image segmentation
+e segmentation accuracy was measured using the truepositive rate true negative rate false positive rate and falsenegative rate by comparing the results from the algorithmwithmanual segmentation results +e practical results of the k-means clustering segmentation algorithm are shown in Table 2
+e practical results of the k-median clustering seg-mentation algorithm are shown in Table 3
+e practical results of the PSO-based segmentationalgorithm are shown in Table 4
+e practical results of the IWPSO segmentation algo-rithm are shown in Table 5
+e practical results of the GCPSO segmentation algo-rithm are shown in Table 6
+e graphical view of the comparison of the true positiverate true negative rate false positive rate and false negativerate for the algorithms used is shown in Figures 6ndash9 It isproved that the true positive and true negative rates are highand false positive and false negative rates are low for theGCPSO algorithm
+e comparative evaluation based on the accuracy of thesegmentation is shown in Table 7 and Figure 10 +e resultsindicate that the GCPSO-based technique has the highestaverage value of accuracy than the other methods
+e resultant images after preprocessing are shown inFigures 11(a) and 11(b)
+e resultant images after segmentation using k-meansclustering are shown in Figure 12
+e resultant images after segmentation using k-medianclustering are shown in Figure 13
+e resultant images after segmentation using the PSOalgorithm are shown in Figure 14
+e resultant images after segmentation using theIWPSO algorithm are shown in Figure 15
+e resultant images after segmentation using theGCPSO algorithm are shown in Figure 16
In an earlier research lung cancer detection was doneusing PSO genetic optimization and SVM algorithm withthe Gabor filter and produced an accuracy of 895 [18]+emethod to detect lung cancer by means of K-NN classifi-cation using the genetic algorithm produced a maximumaccuracy of 90 [19]+e comparative results with respect tothe above-said methods are shown in Table 8
+e graphical comparative analysis between the used andexisting methods is shown in Figure 17
5 Conclusion
In this study various optimization algorithms have beenevaluated to detect the tumor Medical images often need
preprocessing before being subjected to statistical analysis+e adaptive median filter has better results than medianand mean filters because the speckle suppression index andspeckle and mean preservation index values are lower for theadaptive median filter Comparing the five algorithms theaccuracy of the tumor extraction is improved in GCPSOwith the highest accuracy of 958079 and it obtained above90 of precision in all the 20 images It is more accuratewhen compared to the previous method which had an ac-curacy of 90 in 4 out of 10 datasets only In future studiesthe use of more number of optimization algorithms will beincluded to improve the accuracy
Data Availability
+eCTimages data used to support the findings of this studyhave been deposited in the LungCT-Diagnosis repository(doiorg107937K9TCIA2015A6V7JIWX)
Conflicts of Interest
+e authors declare that there are no conflicts of interestregarding the publication of this paper
References
[1] A A Brindha S Indirani and A Srinivasan ldquoLung cancerdetection using SVM algorithm and optimization techniquesrdquoJournal of Chemical and Pharmaceutical Sciences vol 9 no 42016
[2] M Kurkure and A +akare ldquoIntroducing automated systemfor lung cancer detection using Evolutionary ApproachrdquoInternational Journal of Engineering and Computer Sciencevol 5 no 5 pp 16736ndash16739 2016
[3] B Rani A K Goel and R Kaur ldquoA modified approach forlung cancer detection using bacterial forging optimization
868890929496
PSO GA and SVM algorithm
K-NNclassification
using GA
Proposed GCPSOmethod
8950 90
9581
Acc
urac
y
Various algorithms
Comparison between existing and projected methods
Figure 17 Graphical view of accuracy
Table 8 Comparative analysis of accuracy of the projected methodwith various methods
Various methods Accuracy ()PSO GA and SVM algorithm [18] 8950K-NN classification using GA [19] 9000Projected GCPSO method 9581
Computational and Mathematical Methods in Medicine 15
algorithmrdquo International Journal of Scientific Research En-gineering and Technology vol 5 no 1 2016
[4] N Panpaliya N Tadas S Bobade R Aglawe andA Gudadhe ldquoA survey on early detection and prediction oflung cancerrdquo International Journal of Computer Science andMobile Computing vol 4 no 1 pp 175ndash184 2015
[5] G Gupta ldquoAlgorithm for image processing using improvedmedian filter and comparison of mean median and improvedmedian filterrdquo International Journal of Soft Computing andEngineering vol 1 no 5 2011
[6] Tutorial point with digital image processing httpwwwtutorialspointcomdiphistogram_equalizationhtm
[7] K Venkatalakshmi and S M Shalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and K-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Melbourne VIC Australia December2005
[8] K Venkatalakshmi and S M Shalinie ldquoMultispectral imageclassification using modified k-means clusteringrdquo In-ternational Journal on Neural and Mass-Parallel Computingand Information Systems vol 17 no 2 pp 113ndash120 2007
[9] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercy Shalinie ldquoMultispectral image clustering using en-hanced genetic k-means algorithmrdquo Information TechnologyJournal vol 6 no 4 pp 554ndash560 2007
[10] P I Dalatu ldquoTime complexity of k-means and k-mediansclustering algorithms in outliers detectionrdquo Global Journal ofPure and Applied Mathematics vol 12 no 5 pp 4405ndash44182016
[11] K Senthilkumar K Venkatalakshmi K Karthikeyan andP Kathirkamasundari ldquoAn efficient method for segmentingdigital image using a hybrid model of particle swarm opti-mization and artificial bee colony algorithmrdquo InternationalJournal of Applied Engineering Research vol 10 pp 444ndash4492015
[12] K Venkatalakshmi and S Mercyshalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and k-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Bangalore India December 2005
[13] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercyshalinie ldquoA customized particle swarm optimizationfor classification of multispectral imagery based on featurefusionrdquo International Arab Journal of Information Technologyvol 5 no 4 pp 327ndash333 2008
[14] J C Bansal and P K Singh ldquoInertia weight strategies inparticle swarm optimizationrdquo in Proceedings of IEEE WorldCongress on Nature and Biologically Inspired Computing pp640ndash647 Salamanca Spain October 2011
[15] P K Patel V Sharma and K Gupta ldquoGuaranteed conver-gence particle swarm optimization using Personal Bestrdquo In-ternational Journal of Computer Applications vol 73 no 7pp 6ndash10 2013
[16] X Wang L Ge and X Li ldquoEvaluation of filters for Envi-satAsar speckle suppression in pasture areardquo ISPRS Annals ofthe Photogrammetry Remote Sensing and Spatial InformationSciences vol 1ndash7 pp 341ndash346 2012
[17] H Nagaveena D Devaraj and S C Prasanna KumarldquoVessels segmentation in diabetic retinopathy by adaptivemedian thresholdingrdquo International Journal of Science andTechnology vol 1 no 1 pp 17ndash22 2013
[18] A Asuntha N Singh and A Srinivasan ldquoPSO genetic op-timization and SVM algorithm used for lung cancer
detectionrdquo Journal of Chemical and Pharmaceutical Researchvol 8 no 6 pp 351ndash359 2016
[19] P Bhuvaneswari and A Brintha+erese ldquoDetection of cancerin lung with K-NN classification using genetic algorithmrdquoInternational Conference on Nanomaterials and Technologiesvol 10 pp 433ndash440 2014
16 Computational and Mathematical Methods in Medicine
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom
algorithmrdquo International Journal of Scientific Research En-gineering and Technology vol 5 no 1 2016
[4] N Panpaliya N Tadas S Bobade R Aglawe andA Gudadhe ldquoA survey on early detection and prediction oflung cancerrdquo International Journal of Computer Science andMobile Computing vol 4 no 1 pp 175ndash184 2015
[5] G Gupta ldquoAlgorithm for image processing using improvedmedian filter and comparison of mean median and improvedmedian filterrdquo International Journal of Soft Computing andEngineering vol 1 no 5 2011
[6] Tutorial point with digital image processing httpwwwtutorialspointcomdiphistogram_equalizationhtm
[7] K Venkatalakshmi and S M Shalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and K-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Melbourne VIC Australia December2005
[8] K Venkatalakshmi and S M Shalinie ldquoMultispectral imageclassification using modified k-means clusteringrdquo In-ternational Journal on Neural and Mass-Parallel Computingand Information Systems vol 17 no 2 pp 113ndash120 2007
[9] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercy Shalinie ldquoMultispectral image clustering using en-hanced genetic k-means algorithmrdquo Information TechnologyJournal vol 6 no 4 pp 554ndash560 2007
[10] P I Dalatu ldquoTime complexity of k-means and k-mediansclustering algorithms in outliers detectionrdquo Global Journal ofPure and Applied Mathematics vol 12 no 5 pp 4405ndash44182016
[11] K Senthilkumar K Venkatalakshmi K Karthikeyan andP Kathirkamasundari ldquoAn efficient method for segmentingdigital image using a hybrid model of particle swarm opti-mization and artificial bee colony algorithmrdquo InternationalJournal of Applied Engineering Research vol 10 pp 444ndash4492015
[12] K Venkatalakshmi and S Mercyshalinie ldquoClassification ofmultispectral images using support vector machines based onPSO and k-means clusteringrdquo in Proceedings of IEEE In-ternational Conference on Intelligent Sensing and InformationProcessing pp 127ndash133 Bangalore India December 2005
[13] K Venkatalakshmi P Anisha Praisy R Maragathavalli andS Mercyshalinie ldquoA customized particle swarm optimizationfor classification of multispectral imagery based on featurefusionrdquo International Arab Journal of Information Technologyvol 5 no 4 pp 327ndash333 2008
[14] J C Bansal and P K Singh ldquoInertia weight strategies inparticle swarm optimizationrdquo in Proceedings of IEEE WorldCongress on Nature and Biologically Inspired Computing pp640ndash647 Salamanca Spain October 2011
[15] P K Patel V Sharma and K Gupta ldquoGuaranteed conver-gence particle swarm optimization using Personal Bestrdquo In-ternational Journal of Computer Applications vol 73 no 7pp 6ndash10 2013
[16] X Wang L Ge and X Li ldquoEvaluation of filters for Envi-satAsar speckle suppression in pasture areardquo ISPRS Annals ofthe Photogrammetry Remote Sensing and Spatial InformationSciences vol 1ndash7 pp 341ndash346 2012
[17] H Nagaveena D Devaraj and S C Prasanna KumarldquoVessels segmentation in diabetic retinopathy by adaptivemedian thresholdingrdquo International Journal of Science andTechnology vol 1 no 1 pp 17ndash22 2013
[18] A Asuntha N Singh and A Srinivasan ldquoPSO genetic op-timization and SVM algorithm used for lung cancer
detectionrdquo Journal of Chemical and Pharmaceutical Researchvol 8 no 6 pp 351ndash359 2016
[19] P Bhuvaneswari and A Brintha+erese ldquoDetection of cancerin lung with K-NN classification using genetic algorithmrdquoInternational Conference on Nanomaterials and Technologiesvol 10 pp 433ndash440 2014
16 Computational and Mathematical Methods in Medicine
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom
Stem Cells International
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
MEDIATORSINFLAMMATION
of
EndocrinologyInternational Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Disease Markers
Hindawiwwwhindawicom Volume 2018
BioMed Research International
OncologyJournal of
Hindawiwwwhindawicom Volume 2013
Hindawiwwwhindawicom Volume 2018
Oxidative Medicine and Cellular Longevity
Hindawiwwwhindawicom Volume 2018
PPAR Research
Hindawi Publishing Corporation httpwwwhindawicom Volume 2013Hindawiwwwhindawicom
The Scientific World Journal
Volume 2018
Immunology ResearchHindawiwwwhindawicom Volume 2018
Journal of
ObesityJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Computational and Mathematical Methods in Medicine
Hindawiwwwhindawicom Volume 2018
Behavioural Neurology
OphthalmologyJournal of
Hindawiwwwhindawicom Volume 2018
Diabetes ResearchJournal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Research and TreatmentAIDS
Hindawiwwwhindawicom Volume 2018
Gastroenterology Research and Practice
Hindawiwwwhindawicom Volume 2018
Parkinsonrsquos Disease
Evidence-Based Complementary andAlternative Medicine
Volume 2018Hindawiwwwhindawicom
Submit your manuscripts atwwwhindawicom