automated hierarchical posterior map method (hpmm) for ...cikitusi.com/gallery/40-april-691.pdf ·...

Automated Hierarchical Posterior Map Method

(HPMM) for Brain Tumor Detection and

Segmentation using MRI Images C. Jaspin Jeba Sheela

#1, G. Suganthi

*2

# Reg.No. 17221282162010,Research Scholar, St.Xavier’s Autonomous College, Palayamkottai affiliated to Manonmaniam

Sundaranar University, Abishekapatti, Tirunelveli 627012, Tamil Nadu, India

Associate Professor, Department of Computer Science, Women’s Christian College, Nagercoil affiliated to Manonmaniam

Sundaranar University, Abishekapatti, Tirunelveli 627012, Tamil Nadu, India

1E-mail: [email protected]

2E-mail: [email protected]

Abstract— Image segmentation denotes a process of partitioning an image into distinct regions. A large variety of different

segmentation approaches the images which have been developed. Among them, the clustering based approach uses a Hierarchal

Posterior Map Method (HPMM) is proposed for Medical Image Segmentation. This paper describes the segmentation method that

consists of various phases. In the first phase, the MRI brain image acquires from preprocessing which is found CSF, Gray Matter,

White Matter, Edema, Enhanced core. In the second Phase, MR image segmentation is to accurately identify the Gaussian

Normalization. The third phase, Enhance the MRI Image by using Otsu’s Thresholding method. The fourth phase the tumor is detected

initial stage. The fifth phase masked tumor detection, to improve the accuracy of segmentation. Finally the proposed Hierarchal

Posterior Map Method (HPMM) to segment the Brain tumor and non-tumor part of the brain.

Keywords— Hierarchical posterior Map, Gray matter, Brain Tumor, White Matter, Image segmentation

I. INTRODUCTION

The tumor is an abnormal growth of cells in the body. The primary brain tumor is relevant to benign

(Non-cancer) brain tumor. It grows within the brain from cells or tissues present in the normal brain. They

may arise from any cell or tissue but most commonly arise from supporting cells (glial cells). It is a group of

abnormal cells that may start in the brain. It is curable. The secondary brain tumor is relevant to malignant

(cancer) brain tumor. The cells or tissues reach another part of the body, they may grow and form another

tumor is called a secondary brain tumor [2]. When the cancer is spread over the brain, it is not curable.

Image segmentation is typically used to identify the boundary lines in tumor images. The goal of

segmentation is to simplify or change the representation of an image and easier to analyze.

The World Health Organization (WHO) has issued a grading for brain tumors in which, Grade I is

least aggressive and grow slowly, the tumor is less than 2cm. This tumor is relatively slow growing cells and

slightly abnormal appearance [1]. Grade II reproduces and affects nearby tissues, this tumor is between 2 to

5 cm. Grade III is the malignant tumor that reproduces cells and affects tissues, this tumor is more than 5cm.

Grade IV tumor is the most malignant tumors which usually reproduces rapidly and affects nearby normal

brain tissue. In this paper, an efficient segmentation technique for brain MRI is proposed to use the

clustering method [8]. The HPMM is used for segmentation of brain MR Image. The rest of this article is

CIKITUSI JOURNAL FOR MULTIDISCIPLINARY RESEARCH

Volume 6, Issue 4, April 2019

ISSN NO: 0975-6876

http://cikitusi.com/327

mailto:[email protected]

organized as follows: Section II discusses the Literature Survey. Section III discusses the proposed MRI

Detection and Segmentation. Section IV discusses the conclusion and Future Challenges.

II. LITERATURE SURVEY

The paper [1] presents a method Hierarchical Self-Organizing Map (HSOM) and Self Organising

Map (SOM) investigated the multispectral Magnetic Resonance (MR) images. The HSOM is composed into

several layers of Self-Organizing Maps (SOMs). The results of SOM is often used from under or over-

segmentation. The multispectral Magnetic Resonance (MR) images combined the concepts of self-

organization and topographic mapping. The segmentation results of the HSOM are compared with SOM and

used the k-means clustering algorithm to get the accurate result.

The author Adriano Pinto, Sérgio Pereira et al. [2] describe the gliomas are the most common and

aggressive primary brain tumors, with a short life expectancy in their highest grade. Magnetic Resonance

Imaging is the most common imaging technique to assess brain tumors. There is a need for automatic and

robust segmentation methods. We propose an automatic hierarchical brain tumor segmentation pipeline

using Extremely Randomized Trees with appearance- and context-based features. Some of these features are

computed over non-linear transformations of the Magnetic Resonance Imaging images.

The paper [3] presents Image segmentation is one of the most important tasks in medical image

analysis. In brain MRI analysis, image segmentation is commonly used for measuring and visualizing the

brain's anatomical structures, for analyzing brain changes, for delineating pathological regions, and for

surgical planning and image-guided interventions. Various segmentation techniques of different accuracy

and degree of complexity have been developed and reported in the literature. we review the most popular

methods commonly used for brain MRI segmentation. The complexity and challenges of the brain MRI

segmentation are first to introduce the basic concepts, MRI pre-processing steps including image

registration, bias field correction, and removal of nonbrain tissue of image segmentation. Finally, after

reviewing different brain MRI segmentation methods, we discuss the validation problem in brain MRI

segmentation.

The paper [4] suggests the conventional method of detection and classification of brain tumor using

medical resonance brain images. Medical Resonance images contain a noise caused by operator

performance which can lead to serious inaccuracies classification. The use of Support Vector Machine, K

Means, and PCA have shown great potential in this field. Principal Component Analysis gives a fast and

accurate tool for Feature Extraction of the tumors. K Means is used for Image Segmentation. Support

Vector Machine with Training and Testing of Brain Tumor Images techniques are implemented for

automated brain tumor classification.



ISSN NO: 0975-6876


The paper [5] describes the implementation of a simple algorithm for automatic brain tumor

segmentation. The brain tumor is commonly diagnosed by Computer tomography and Magnetic

Resonance Imaging in clinical treatment. The paper uses Simple Linear Iterative Clustering (SLIC) to

segment brain images according to their spatial and color proximities. The ratio of the mean and

variance of the image pixels are determined in order to obtain an optimum threshold

value. Region merging after thresholding was carried out. The final output image was an image with

tumor sections circled.

G. Evelin Suji et al. [6] describes an efficient image segmentation technique that can distinguish

the pathological tissues such as edema and tumor from the normal tissues such as White Matter (WM),

Grey Matter (GM), and Cerebrospinal Fluid (CSF). Thresholding is simpler and most commonly used

techniques in image segmentation. This technique can be used to detect the contour of the tumor in the

brain.

The author Swathi P. S et al. [7] proposes the segmentation of brain tumors in Magnetic Resonance

Images (MRI) is difficult to find shapes, locations, image intensities. It is projected to encapsulate and

evaluate the method of mechanical recognition of brain tumor through Magnetic Resonance Image (MRI)

with Histogram Thresholding and Artificial Neural Network. The anticipated method can be effectively

useful to distinguish the shape of the tumor and its geometrical measurement. Also in this paper, a tailored

Artificial Neural Network (ANN) model that is based on learning vector quantization with image and data

analysis and exploitation technique is anticipated to carry out a computerized brain tumor classification

using MRI-scans. The estimation of the adapted ANN classifier is deliberate in terms of the guidance

performance, classification accuracies and computational time. This paper presents two techniques for the

exposure purpose; the first one is Histogram Thresholding and another is an Artificial Neural Network

technique. The planned Neural Network technique consists of feature extraction, dimensionality,

recognition, segmentation, and organization.

The paper [8] presents the segmentation and identification of brain tumor from MR images

efficiently. The segmentation process is extracted different tumor tissues such as active, tumor, necrosis,

and edema from the normal brain tissues such as white matter (WM), gray matter (GM), as well

cerebrospinal fluid (CSF). The brain tumors most of the time detected easily from brain MR image but

required the level of accuracy, reproducible segmentation, abnormalities classification are not predictable

and straightforward. The segmentation of brain tumor is composed of many stages. The manual

segmentation of brain MR images is very time consumption and tedious task, and hence it is associated



ISSN NO: 0975-6876


with many challenges. Therefore, we need an automated segmentation method for brain images. There are

many techniques presented to investigate the performance of automated computerized brain tumor

detection for the medical analysis purpose.

III. PROPOSED MRI DETECTION AND SEGMENTATION

Hierarchical Posterior Map organizing the inner portion of the human body for medical diagnosis.

HPMM classifies the inner layer of the image has to identify the tumor and non-tumor area [9]. The

posterior map shows the x and y-direction of the tumor. The combination of Gaussian convolution (GC)

and Gaussian Normalization (GN) introduce a Gaussian Weighted image (GW) The lowest level weight

vector is achieved by the abstraction level. We have also achieved a higher value of tumor pixels by this

HPMM approach [2]. The multilayer segmentation results of the HPMM are shown to have interesting

consequences from the viewpoint of clinical diagnosis.

Gradient helps to identify the initial mask of the tumor, identify the flow of direction and identify

the shape of the tumor. The level set of the gradient is dynamic and spread highly. The gradient is one of

the tools to identify the extract seed point of the tumor. The gradient is used to preprocess a grayscale

image. The gradient may be used to identify the internal and external Neighbourhood direction of the

tumor image. The morphological concept is also used in the gradient. The result of the gradient is much-

improved segmentation [3].

Gray matter has a gray color in the living brain which contains the cell bodies, dendrites and axon

terminals of neurons. It is made up of nerve cell bodies and composed by neurons. It is an intermediate

signal intensity [10]. White Matter is made of axons connecting different parts of grey matter. White

matter occurs in both the brain and the spinal cord. White matter contains relatively few cell bodies and is

composed of a long-range myelinated axon. The color difference may arise mainly the whiteness of

myelin. It is a high signal intensity.CSF stands for cerebrospinal fluid [11]. It is a colorless fluid from

the brain and it is a clear substance that circulates through the brain. CSF fluid gets absorbed and it is

produced from the choroid plexus. CSF is located in Ventricles. A brain tumor may cause and build up or

blockage of CSF. A small amount of fluid is produced by ependymal cells. Normalization shows the

intensity. It helps to find out the neighborhood value and combines the row and column.In this

normalization, we use smoothing and gradient to enhance the tumor image and remove all the errors.

When smoothing the tumor shape will be improved. The smoothed image was caused by the blurring of a

boundary. The result of smoothing may remove the noise or errors in the tumor image. when we use the

smoothing image as an average mask of size. It helps to identify the tumor successfully. Region growing



ISSN NO: 0975-6876


is a procedure that groups the pixels of Tumor image. The basic approach is to start with a set of seed

points and from these grow regions by appending to each seed or neighboring pixels. It helps to identify

the centroid of seed. Region-Based Segmentation is a partition of an image into region[17]. The

boundaries between region are based on the intensity level.

Segmentation techniques that are based on finding the regions directly. The procedure is used to

identify the region growing in a set of seed points [20]. The region is merged or split to satisfy the

condition or purpose. The whole seed point of the tumor may be splitting and merging. During splitting

the region should be split and merging is a process to find out good segmentation result. The segmentation

techniques are used to find out the extraction of the tumor region in the MRI image. The image

segmentation refers to partitioning the image into multiple regions. Extraction may be segmenting the

region accurately.

A. Thresholding

Thresholding is one of the frequently used methods for Image Segmentation. The simplest

method of image segmentation is called Thresholding method [19][21]. The value is true to grayscale

image into a binary image. The key for this method is to select the Threshold value. This value is defined

as Mathematically [5][12]. T represents the threshold value. Let f (x,y) be the input image, G (x,y) is the

segmented output image. Equation 1 shows the Threshold binary image g (x,y) is defined as

(1)

Let f(x,y) be the Input image, T Threshold Value , G(x,y) is the segmented Image. Equation 2 shows the

two threshold values and this equation segments the image into three groups.

G(x,y) = (2)

The parameter in image segmentation using thresholding technique is the choice of selecting threshold

value T. Thresholding operation is defined as: T = M [x, y, m(a, b), n(a, b)] In given equation , T stands

for the threshold; m (a, b) is the gray value of point (a, b) and b(a, b) denotes some local property of the

point such as the average gray value of the neighborhood centered on point (a, b).There are two types of

thresholding methods. 1) Global thresholding: When threshold T depends only on gray-level values i.e.

m(a,b) and the value of T solely relates to the character of pixels, this thresholding technique is called

a, if f(x,y) > T

b, if f(x,y) ≤ T G (x,y) =

=

a, if f(x,y) > T2

b, if T1< f(x,y) ≤ T2

c, if f(x,y) ≤ T1



ISSN NO: 0975-6876


global thresholding. 2) Local thresholding: If threshold T depends on n(a,b) and m(a,b), this thresholding

is called local thresholding. This method divides an original image into several sub regions, and chooses

various thresholds T for each sub region reasonably.

B. Thresholding using OTSU’S Method

During the Otsu’s method may threshold the normalized range. This method may confirm every

dark point on the tumor image and every light point of the tumor image is a thresholding error [6] [7].

So only the segmentation was highly unsuccessful. Generally, image processing an algorithm is capable

of choosing the threshold automatically based on image data [18]. An MRI brain image can be

enhanced using unsharp approach before segmentation of the MRI image using the Otsu method. The

Otsu allows us to detect a suitable threshold to segment the tumor part from the image [13][15].The

steps of the Otsu’s Threshold method is,

1.Compute the Threshold of the Input Image

2. Compute the additive sums K(n) for n= 0,1,2,........L-1

3. Compute the additive means M(n) for n=0,1,2,........L-1

4. Compute the Otsu’s intensity Mean Mo

5. Compute the class between variance σ 2A(n) for n = 0,1,2........L-1

6. Obtain Otsu’s threshold n* as the value of n for which σ2

(n) Maximum.

7. If the minimum is not unique, obtain n* by averaging the values of n corresponding to the various

maxima detected separability measure β2, at n = n*

8. Finally compute the output.

C. Preprocessing

The MRI Images are to be preprocessed before the realization of them for the segmentation process.

Preprocessing of brain MR image may remove the non-tumor brain tissues. Preprocessing will eliminate

errors caused by taking the image and to reduce brightness effects on the Image. Two Techniques are

commonly used before pre-processing and after pre-processing [16]. Before Preprocessing, methods can

find CSF, Gray matter, White Matter, Edema and Enhanced Core. After preprocessing methods identify

the Gaussian Normalization, then enhance the image by using OTSU’s method [4]. After pre-processing

we remove the Artifacts or noise and enhance the image into a better image. The Advantages are to

segment an image accurately, flexible approach, time-consuming etc. The disadvantage of an image is

defined poorly, difficult estimation, misclassification etc



ISSN NO: 0975-6876


T1 Weighted image is dark. The image contrast is based on longitudinal. T1 images are similar to

normal anatomical details. T2 Weighted image is bright. The image contrast is based on transverse. T2

images are similar to pathological details. This shows the signal enhancement after administration of

contract [14]. FLuid Attenuated Inversion Recovery (FLAIR). It is similar to the T2 Weighted image.

FLAIR is bright signals. It allows the better detection and good imaging of the tumor. The abnormality of

FLAIR is usually known as edema.

Figure 1 : Block Diagram of the Proposed Hierarchal Posterior Map Method (HPMM)

Segmented

output

Input

Images

Proposed

HPMM

Preprocessing

Gaussian

T1, T2,

T1C, Flair

Otsu’s Thresholding

(Enhance the image)

Initial

Tumor

Detection

Masked

Tumor

Detection

Segmented Output

Masked Tumor Detection

Intial Tumor Detection

Tumor

Non Tumor

Convolutional

Normalization

Weighted Images

After Preprocessing

Normailzation

Weighted vote aggregation

CSF

Gray

Matter

White

Matter

Edema

Enhanced

Core

CSF

GM

WM

Edema

EC



ISSN NO: 0975-6876


Algorithm: Hierarchal Posterior Map Method (HPMM)

Input : MRI Images

Output : Detected and Segmented output

Step 1: Read the MRI Image

Step 2: Processing techniques are used to pre-process the image we get CSF, Gray Matter, White Matter,

Edema and Enhancing Core.

Step 3: Estimate the Gradient and identify the X, Y direction of the Tumor in the image.

Step 4: To apply the Gaussian Normalization (GN) and Gaussian Convolution (GV)

Step 5: Gaussian Normalization may smoothing the image and improve the shape of the image.

Step 6: Combine GV and GN then finding the Gaussian weighted image(GW).

Step 7: Apply thresholding and Otsu’s Method for segmenting the tumor.

Select threshold T

S=Segmentation

Edge E = e1, e2, e3………..en

R = (r1, r2, r3, ……………..rn)

T = {(Sr)/ r=1,2, ……..n} set of region for segmentation

For each r1 € E do

R1 Ej € S (Random selection of edge in segmentation)

end

Step 8 : To detect the Initial tumor

Step 9 : Masked Tumor detection Method remove the artifacts in the image.

Step 10: To apply the (HPMM) Hierarchal Posterior Map method for detecting and Segmenting the tumor.

Step 11: Finally the HPMM Tumor and Non-Tumor Segmented.

IV. RESULTS AND DISCUSSION

Figure 2 . Illustrates the Input Images for BRAT’S database

(a) T1 Image (b) T2 Image (c) T1C (d) FLAIR

(a) T1

(b) T2

(c) T1C

(d) FLAIR

CSF

(a) CSF

GM

(b)Gray Matter

WM

(c) White

Matter

Edema

(d) Edema

EC

(e)Enhancing

Core (EC)



ISSN NO: 0975-6876


Figure 3. preprocessing stages (a) CSF (b) Gray matter (c) White matter (d) Edema

(d) Enhancing Core (EC).

Table 1. Analysis of Low and High grade cases using Tumor and non Tumor

Cases Tumor Non

Tumor

TP TN FN FP

Low Grade 13 8 8 6 5 2

High Grade 51 36 40 22 11 14

Figure 4: Analysis chart of Low and High grade cases using Tumor and non Tumor

During performance evaluation, there exist some performance evaluation cases for each

implemented technique. These cases have described below:

True Positive (TP)

If one detect the condition when condition present there, the test result is called true

positive.

True Negative (TN)

If one does not detect the condition when condition is not present there, the test result is

called true negative.

False Positive (FP)

If one does not detect the condition when condition is present there, the test result is called

false positive.

False Negative (FN)

If one detects the condition when condition is not present there, the test result is called false

negative.

False Positive (FP)

If one detects the condition when condition is present there but do not detect it correctly,

then the test result is called false positive.

Table 1 shows, there are 13 cases in Low Grade Tumor, 8 cases in Low Grade Non Tumor,

51 cases in High Grade Tumor and 36 cases High Grade Non Tumor. The Tumor values are



ISSN NO: 0975-6876


calculated by Truth Positive (TP) and False Negative (FN). The Non Tumor values are calculated

by Truth Negative (TN) and False Positive (FP).

Fig 4 Analysis chart we find the Low Grade and High Grade Tumor and Non Tumor case

lists. Then the Truth Positive (TP), Truth Negative (TN), False Positive (FP), False Negative (FN)

values are calculated.

Table 2: Comparison of Low and High grade Tumor using various methods

Table 2 combines the Low Grade and High Grade tumors .In this table we calculate the

values of JSI , Dice, Sensitivity, Specificity, positive prediction value (PPV), Euclidean Distance

(ED) and we have combine the Low Grades and high Grades of above 6 methods.

0

1

2

3

4

5

Low Grade High Grade Combined

JSI

Dice

Sensitivity

Specificity

PPV

ED

Figure 5: Analysis Chart to compare various methods

The chart actually has 6 Methods. In this chart we have clearly identified the combination

of Low Grade and High Grades of 6 methods. JSI, Dice, Sensitivity, Specificity, Positive

Predictive Value (PPV) ,Euclidean Distance (ED).

Method Low

Grade

High Grade Combined

JSI 0.533 0.6153 1.1483

Dice 0.695 0.761 1.456

Sensitivity 0.615 0.784 1.399

Specificity 0.750 0.611 1.361

PPV 0.800 0.7407 1.540

ED 1.300 3.333 4.633



ISSN NO: 0975-6876


Figure 6: Performance comparision of Brain Tumor Segmentation using BRAT’S 2015 databases

using 3 methods (a) T1 input Image using Gaussian Convolution (b) T1 input image using Gaussian

Normalization (c) T1 input image using Weighted Image. (d) T2 input Image using Gaussian

Convolution (e) T2 input image using Gaussian Normalization (f) T2 input image using Weighted

Input

Images

Gaussian

Convolution

Gaussian

Normalization

Gaussian Weighted

Image

T1

Guassian Convolution

(a)

(b)

(c )

T2

Guassian Convolution

(d)

Normailzation

(e)

7 class Intensity Likelihoods

(f)

T1C

T1-Guassian Convolution TIC-Normalization FLAIR-Normalization T2-Guassian Convolution

Weighted vote aggregation 7 class Intensity Likelihoods CSF GM

WM Edema EC Intial Maximium Posterior Map

Intial Tumor Detection Masked Tumor Detection Segmented Output

0 200 400 600 800 1000

0.7

0.8

0.9

1

Dic

e I

ndex

Atlas Selection count

Accuracy of Segmentation

(g)





0 200 400 600 800 1000

0.7

0.8

0.9

1

Dic

e I

ndex



(h)





0 200 400 600 800 1000

0.7

0.8

0.9

1

Dic

e I

ndex



(i)

FLAIR

After Preprocessing

(j)

Normailzation

(k)

Weighted vote aggregation

(l)



ISSN NO: 0975-6876


Image. (g) T1C input Image using Gaussian Convolution (h) T1C input image using Gaussian

Normalization (i) T1C input image using Weighted Image. (j) FLAIR input Image using Gaussian

Convolution (k) FLAIR input image using Gaussian Normalization (l) FLAIR input image using

Weighted Image.

Figure 7 shows the Initial and Masked tumor detection. Initial tumor detection helps to identify the

starting point or seed point of the tumor. The detection may be easily misclassified, so only we identify the

initial or seed point of the tumor. The line shows the X,Y direction of initial tumor, the midpoint of the X

and Y directions show the seed point of the initial tumor. The initial tumor of MRI image is having

Artifacts, during the artifacts or error the image may become blur. Masked tumor helps to remove the

Artifacts or error in the initial tumor image. During masking the image may be a better image or enhanced

image.

Figure 7: Initial and Masked Tumor

Intial Maximium Posterior Map

(a) Hierarical

Posterior Map Method

(HPMM)

Segmented Output

(b) HPMM- Tumor

and non tumor

Segmented output

Figure 8: Proposed segmented HPMM output images

V. CONCLUSIONS

In this article, the automated brain tumor segmentation techniques of MRI have been used. MRI

based brain tumor segmentation due to the good soft tissue contrast and noninvasive of MRI. This paper

Intial Tumor Detection

(a) Initial Tumor

Masked Tumor Detection

(b) MaskedTumor



ISSN NO: 0975-6876


proposes a work on brain tumor detection system based on HPMM. The proposed work includes

preprocessing, Gaussian Normalization (Smoothing the image), Otsu’s Thresholding, Initial Tumor

Detection and Masked tumor detection. Otsu’s Thresholding is used to enhance tumor Image. Gaussian

shows the combination of Gaussian Normalization (GN) and Gaussian Convolution (GC) in to weighted

image. The HMPP may identify the tumor and Non Tumor detection and Segmentation.

REFERENCES

[1]. Suchendra Bhandarkar, P. Nammalwar et al “Segmentation of multispectral MR images using a hierarchical self-organizing map” Computer-Based

Medical Systems, 2001.

[2]. Adriano Pinto, Sérgio Pereira et al ”Hierarchical Brain Tumour Segmentation using Extremely Randomized Trees” Pattern Recognition 82, May 2018

[3].Ivana Despotović, Bart Goossens, and Wilfried Philips, ”MRI Segmentation of the Human Brain: Challenges, Methods, and Applications”

Computational and Mathematical Methods in Medicine” March 2015.

[4]. Arun Kumar et al, “A Novel Approach for Brain Tumor Detection Using Support Vector Machine, K-Means and PCA Algorithm”, International Journal

of Computer Science and Mobile Computing, Vol.4 Issue.8, pg. 457-474, August- 2015.

[5]. Prince Ebenezer Adjei, Henry Nunoo-Mensah, et al “Brain Tumor Segmentation using SLIC Superpixels and Optimized Thresholding Algorithm” ,

Article in International Journal of Computer Applications, October 2018.

[6]. G. Evelin Sujji, Y.V.S. Lakshmi , G. Wiselin Jiji et al “ MRI Brain Image Segmentation based on Thresholding “,International Journal of Advanced

Computer Research (ISSN (print): 2249-7277, Volume-3 Number-1 Issue-8 March-2013.

[7]. Swathi P S, Deepa Devassy et al,” Brain Tumor Detection and Classification Using Histogram Thresholding and ANN”, International Journal of

Computer Science and Information Technologies (IJCSIT), Vol. 6 (1) , 2015, 173-176.

[8]. Ruchi D. Deshmukh, and Prof. Chaya Jadhav, “Study of Different Brain Tumor MRI Image Segmentation Techniques”, International Journal of

Computer Science Engineering and Technology ( IJCSET) , Vol 4, April 2014.

[9]. D.Selvaraj, R.Dhanasekaran et al “MRI Brain Image Segmentation Techniques - A Review”, Indian Journal of Computer Science and Engineering

(IJCSE) ISSN : 0976-5166, Vol. 4 No.5 Oct-Nov 2013.

[10]. T. Logeswari1 and M. Karnan2 “An improved implementation of brain tumor detection using segmentation based on soft computing”, Journal of

Cancer Research and Experimental Oncology, Vol. 2, pp. 006-014, March, 2010.

[11]. Ashraf M. Said, Fatma S. Ibrahim et al, “Comparative Study of Segmentation Techniques for Detection of Tumors Based on MRI Brain Images”

September 4, 2017.

[12]. Aaron Gonsalves, Rhea Machado et al “Comparative Evaluation of Thresholding and Segmentation Algorithms” International Journal of Computer

Science and Information Technologies (IJCSIT) , Vol. 6, 2015.

[13]. Kirti Mittal, Abhishek Shekhar et al,” Brain Tumour Extraction using Otsu Based Threshold Segmentation” , International Journal of Advanced

Research in Computer Science and Software Engineering” Volume 7, Issue 4, April 2017.

[14]. Deepak Agrawal , Vyom kulshreshtha, Dr. Pankaj Sharma, “Brain Tumor Detection Using K-means Clustering and Threshold Segmentation”,

International Journal of Advanced Research in Science, Engineering and Technology Vol. 5, Issue 3 , March 2018.

[15]. V. Prema , M. Sivasubramanian et al , ” Brain Cancer Feature Extraction Using Otsu's Thresholding Segmentation” International Journal of

Computer Application (2250-1797) Volume 6– No.3, June 2016.

[16]. Uma-e-Hani, Saeeda Naz et al “Automated Techniques for Brain Tumor Segmentation and Detection: A Review Study “,Conference Paper · October

2017.

[17]. Shamla Mantri , Aditi Jahagirdar et al “Automated Preliminary Brain Tumor Segmentation Using MRI Images” International Journal of Computer

Science Issues (IJCSI), Vol. 11, Issue 3, No 1, May 2014.

[18]. Chunshi Sha, Jian Hou , Hongxia Cui et al “A robust 2D Otsu’s thresholding method in image segmentation” J. Vis. Commun. Image R. Elsiver, 26

October 2016.



ISSN NO: 0975-6876


https://www.researchgate.net/profile/Suchendra_Bhandarkar

https://www.researchgate.net/scientific-contributions/2032527329_P_Nammalwar

https://www.researchgate.net/profile/Adriano_Pinto3

https://www.researchgate.net/scientific-contributions/2080111955_Sergio_Pereira

https://www.researchgate.net/journal/0031-3203_Pattern_Recognition

https://www.ncbi.nlm.nih.gov/pubmed/?term=Despotovi%26%23x00107%3B%20I%5BAuthor%5D&cauthor=true&cauthor_uid=25945121

https://www.ncbi.nlm.nih.gov/pubmed/?term=Goossens%20B%5BAuthor%5D&cauthor=true&cauthor_uid=25945121

https://www.ncbi.nlm.nih.gov/pubmed/?term=Philips%20W%5BAuthor%5D&cauthor=true&cauthor_uid=25945121

[19]. Muhammad Sheraz Arshad Malik1, Samreen Shabiret et al ,” A Comparative Analysis of K-means, Thresholding and Region Growing Algorithms for

Segmentation of Brain Tumor of MRI Images”International Journal of Computer Science and Network Security (IJCSNS), VOL.18, No.9, September

2018.

[20]. T.Logeswari and M.Karnan ,” An Improved Implementation of Brain Tumor

Detection Using Segmentation Based on Hierarchical Self Organizing Map” International Journal of Computer Theory and Engineering, Vol. 2, No. 4,

August 2010.

[21]. Roopali R.Laddha, S.A.Ladhake “A Review on Brain Tumor Detection Using Segmentation And Threshold Operations” International Journal of

Computer Science and Information Technologies, Vol. 5, 2014.

[22]. Nilakshi Devi , Prasanta Kr. Baruah, Kaustubh Bhattacharyya et al “Brain Tumor Detection Using Soft Computing Tools” International Journal of

Science and Research (IJSR), Index Copernicus Value 2013.



ISSN NO: 0975-6876


automated hierarchical posterior map method (hpmm) for ...cikitusi.com/gallery/40-april-691.pdf ·...

Documents