assessment of internal bruise volume of selected fruits using mr imaging ta-te lin, yu-che cheng,...
TRANSCRIPT
ASSESSMENT OF INTERNAL BRUISE VOLUME ASSESSMENT OF INTERNAL BRUISE VOLUME OF SELECTED FRUITS USING MR IMAGINGOF SELECTED FRUITS USING MR IMAGING
Ta-Te Lin, Yu-Che Cheng, Jen-Fang YuTa-Te Lin, Yu-Che Cheng, Jen-Fang Yu
Department of Bio-Industrial Mechatronics Engineering,Department of Bio-Industrial Mechatronics Engineering,National Taiwan University,National Taiwan University,
Taipei, Taiwan, ROCTaipei, Taiwan, ROC
INTRODUCTIONINTRODUCTION
MR ImagingMR Imaging Internal Bruise EvaluationInternal Bruise Evaluation Bruise DetectionBruise Detection 3D Reconstruction3D Reconstruction
OBJECTIVESOBJECTIVES
To develop a measurement method based To develop a measurement method based on MR imaging to compute the internal on MR imaging to compute the internal bruise volume of selected fruits.bruise volume of selected fruits.
To develop image segmentation methods To develop image segmentation methods for bruise detection in MR images.for bruise detection in MR images.
To compare bruise volume estimation using To compare bruise volume estimation using conventional methods and the MR imaging conventional methods and the MR imaging method.method.
MATERIALS & METHODSMATERIALS & METHODS
MR image acquisitionMR image acquisition Impact bruiseImpact bruise Image segmentation of bruise regionImage segmentation of bruise region Volume estimationVolume estimation
MAJOR PARAMETERS IN MR IMAGINGMAJOR PARAMETERS IN MR IMAGING
MR IMAGE ACQUISITIONMR IMAGE ACQUISITION
TRTR TETE NEXNEX FOV FOV ResolutionResolution Slice thicknessSlice thickness
BRUKER S330 MR SCANNER (3.0 T)BRUKER S330 MR SCANNER (3.0 T)
MR IMAGE ACQUISITIONMR IMAGE ACQUISITION
MR IMAGES OF SELECTED FRUITSMR IMAGES OF SELECTED FRUITS
MR IMAGE ACQUISITIONMR IMAGE ACQUISITION
AppleTR = 4000 msTE = 45 ms
PeachTR = 5000 msTE = 60 ms
MangoTR = 4500 msTE = 60 ms
PlumTR = 5000 msTE = 60 ms
DROP TESTDROP TEST
IMPACT BRUISEIMPACT BRUISE
h1
h2
fruit
m
)( 21 hhmgE
BRUISE VOLUME ESTIMATIONBRUISE VOLUME ESTIMATION
IMPACT BRUISEIMPACT BRUISE
Plan View
Side section view
FULL DEPTH METHODFULL DEPTH METHOD
IMPACT BRUISEIMPACT BRUISE
)43(24
221 b
b dwd
V
ELLIPSOID METHODELLIPSOID METHOD
IMPACT BRUISEIMPACT BRUISE
8
)(
3
4 21wwddV tb
IMAGE SEGMENTATIONIMAGE SEGMENTATION
Manual thresholdManual threshold
Automatic threshold Automatic threshold Successive-iteration methodSuccessive-iteration method Kapur et al. methodKapur et al. method Moment preserving method Moment preserving method
IMAGE SEGMENTATIONIMAGE SEGMENTATION
BRUISE REGIONBRUISE REGION
IMAGE SEGMENTATIONIMAGE SEGMENTATION
AUTOMATIC THRESHOLD METHODSAUTOMATIC THRESHOLD METHODS
Successive-iteration method
Kapur et al. method
Moment preserving method
RESULTSRESULTS
MR images of internal bruiseMR images of internal bruise Comparisons of estimation methodsComparisons of estimation methods Image segmentation Image segmentation Bruise volume vs. impact energyBruise volume vs. impact energy 3D reconstruction of bruise volume3D reconstruction of bruise volume
MR IMAGES OF INTERNAL BRUISEMR IMAGES OF INTERNAL BRUISE
APPLEAPPLE
MR IMAGES OF INTERNAL BRUISEMR IMAGES OF INTERNAL BRUISE
PLUMPLUM
COMPARISONS OF ESTIMATION METHODSCOMPARISONS OF ESTIMATION METHODS
R2 = 0.5813
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
20000
0 5000 10000 15000 20000
Volume by MRI (mm³)
Estim
ate
d V
olu
me (
mm
³)
R2 = 0.5813
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
20000
0 5000 10000 15000 20000
Volume by MRI (mm³)
Estim
ate
d V
olu
me (
mm
³)
Full Depth Method vs. MRI MethodFull Depth Method vs. MRI Method
APPLESAPPLES
COMPARISONS OF ESTIMATION METHODSCOMPARISONS OF ESTIMATION METHODS
Ellipsoid Method vs. MRI MethodEllipsoid Method vs. MRI Method
R2 = 0.6242
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
20000
0 5000 10000 15000 20000
Volume by MRI (mm³)
Es
tim
ate
d V
olu
me
(m
m³)
R2 = 0.6242
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
20000
0 5000 10000 15000 20000
Volume by MRI (mm³)
Es
tim
ate
d V
olu
me
(m
m³)
APPLESAPPLES
R2 = 0.8583
0
2000
4000
6000
8000
10000
12000
0 2000 4000 6000 8000 10000 12000
Volume by MRI (mm³)
Estim
ate
d V
olu
me (
mm
³)
R2 = 0.8583
0
2000
4000
6000
8000
10000
12000
0 2000 4000 6000 8000 10000 12000
Volume by MRI (mm³)
Estim
ate
d V
olu
me (
mm
³)
COMPARISONS OF ESTIMATION METHODSCOMPARISONS OF ESTIMATION METHODS
Full Depth Method vs. MRI MethodFull Depth Method vs. MRI Method
PLUMSPLUMS
R2 = 0.7858
0
2000
4000
6000
8000
10000
12000
0 2000 4000 6000 8000 10000 12000
Volume by MRI (mm³)
Est
imat
ed V
olum
e (m
m³)
R2 = 0.7858
0
2000
4000
6000
8000
10000
12000
0 2000 4000 6000 8000 10000 12000
Volume by MRI (mm³)
Est
imat
ed V
olum
e (m
m³)
COMPARISONS OF ESTIMATION METHODSCOMPARISONS OF ESTIMATION METHODS
Ellipsoid Method vs. MRI MethodEllipsoid Method vs. MRI Method
PLUMSPLUMS
SUMMARY TABLESUMMARY TABLE
COMPARISONS OF ESTIMATION METHODSCOMPARISONS OF ESTIMATION METHODS
Method Apple
Slope Intercept R2 Relative Error(%)
RMSE (mm3)
Full Depth 1.03 -50.92 0.58 25.8±19.1 2602
Ellipsoid 1.06 214.9 0.62 25.8±22.4 2547
Method Plum
Slope Intercept R2 Relative Error(%)
RMSE (mm3)
Full Depth 1.67 248.0 0.89 90.7±69.9 2556
Ellipsoid 1.39 -1256.5 0.79 53.2±40.9 1280
IMAGE SEGMENTATIONIMAGE SEGMENTATION
R2 = 0.8556
0
2000
4000
6000
8000
10000
12000
0 2000 4000 6000 8000 10000 12000
Volume by MRI (mm³)
Est
imat
ed V
olum
e (m
m³)
R2 = 0.8556
0
2000
4000
6000
8000
10000
12000
0 2000 4000 6000 8000 10000 12000
Volume by MRI (mm³)
Est
imat
ed V
olum
e (m
m³)
Successive-iteration MethodSuccessive-iteration Method
IMAGE SEGMENTATIONIMAGE SEGMENTATION
Kapur et al. MethodKapur et al. Method
R2 = 0.9114
0
2000
4000
6000
8000
10000
12000
0 2000 4000 6000 8000 10000 12000
Volume by MRI (mm³)
Est
imat
ed V
olum
e (m
m³)
R2 = 0.9114
0
2000
4000
6000
8000
10000
12000
0 2000 4000 6000 8000 10000 12000
Volume by MRI (mm³)
Est
imat
ed V
olum
e (m
m³)
IMAGE SEGMENTATIONIMAGE SEGMENTATION
R2 = 0.8556
0
2000
4000
6000
8000
10000
12000
0 2000 4000 6000 8000 10000 12000
Volume by MRI (mm³)
Est
imat
ed V
olum
e (m
m³)
R2 = 0.8556
0
2000
4000
6000
8000
10000
12000
0 2000 4000 6000 8000 10000 12000
Volume by MRI (mm³)
Est
imat
ed V
olum
e (m
m³)
Moment Preserving MethodMoment Preserving Method
IMAGE SEGMENTATIONIMAGE SEGMENTATION
SUMMARY TABLESUMMARY TABLE
Method Plum
Slope Intercept R2 Relative error
(%)RMSE(mm3)
Successive iteration
1.17 381.78 0.86 40.8±42.5 1143
Kapur et al. 1.71 -983.55 0.91 31.1±25.2 1691
Moment preserving 1.35 328.09 0.89 54.6±36.0 1581
BRUISE VOLUME VS IMPACT ENERGYBRUISE VOLUME VS IMPACT ENERGY
APPLESAPPLES
R2 = 0.4318
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
20000
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
Energy (J)
Bruis
e Volu
me (m
m3 )
R2 = 0.4318
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
20000
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
Energy (J)
Bruis
e Volu
me (m
m3 )
R2 = 0.7304
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
20000
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
Energy (J)Br
uise
Vol
ume
(mm
³)
Ellipsoid Method MR Imaging Method
BRUISE VOLUME VS IMPACT ENERGYBRUISE VOLUME VS IMPACT ENERGY
PLUMSPLUMS
Ellipsoid Method MR Imaging Method
R2 = 0.3506
0
2000
4000
6000
8000
10000
12000
0.0 0.2 0.4 0.6 0.8 1.0
Energy (J)
Bruis
e Volu
me (m
m³)
R2 = 0.3506
0
2000
4000
6000
8000
10000
12000
0.0 0.2 0.4 0.6 0.8 1.0
Energy (J)
Bruis
e Volu
me (m
m³)
R2 = 0.5873
0
2000
4000
6000
8000
10000
12000
0.0 0.2 0.4 0.6 0.8 1.0
Energy (J)Es
timate
d Volu
me (m
m³)
3D RECONSTRUCTION3D RECONSTRUCTION
Apple Plum
Peach Mango
3D RECONSTRUCTION3D RECONSTRUCTION
Full DepthFull Depth EllipsoidEllipsoid
CONCLUSIONSCONCLUSIONS
The MR imaging approach provides an more accurate The MR imaging approach provides an more accurate method to access internal bruise of selected fruits.method to access internal bruise of selected fruits.
The volume estimation errors for apples were 25.8The volume estimation errors for apples were 25.8±±15.6% 15.6% and 22.3and 22.3±±12.3% using full depth method and ellipsoid 12.3% using full depth method and ellipsoid method, respectively. For plums, the estimation error were method, respectively. For plums, the estimation error were 90.790.7±±15.6% and 70.015.6% and 70.0±±12.3%, respectively.12.3%, respectively.
The three automatic threshold methods tended to over The three automatic threshold methods tended to over estimate bruise area. For MR images of plums, the best estimate bruise area. For MR images of plums, the best method were successive-iteration method which yielded an method were successive-iteration method which yielded an estimation error of 40.8±42.5%. estimation error of 40.8±42.5%.
The 3D-reconstructed bruise volume helps in visualizing the The 3D-reconstructed bruise volume helps in visualizing the extent and modes of impact bruise for selected fruits.extent and modes of impact bruise for selected fruits.
THANK YOUTHANK YOU 謝 謝謝 謝