An enhanced liver stages classification in 3 d ct and 3d-us images using glrlm and 3d-cnn

http://www.iaeme.com/IJEET/index.asp 171 editor@iaeme.com
International Journal of Electrical Engineering and Technology (IJEET)
Volume 12, Issue 2, February 2021, pp. 171-184, Article ID: IJEET_12_02_016
Available online at http://www.iaeme.com/ijeet/issues.asp?JType=IJEET&VType=12&IType=2
ISSN Print: 0976-6545 and ISSN Online: 0976-6553
DOI: 10.34218/IJEET.12.2.2021.016
© IAEME Publication Scopus Indexed
AN ENHANCED LIVER STAGES
CLASSIFICATION IN 3D-CT AND 3D-US
IMAGES USING GLRLM AND 3D-CNN
A. Bathsheba Parimala and R. S. Shanmugasundaram
Vinayaka Mission Research Foundation, Deemed to be University, Salem, Tamil Nadu, India
ABSTRACT
Ultrasound (US) and computed tomography (CT) plays a fundamental role in the
classification and staging of liver infections. But the most popular technologies 3D-US
and 3D-CT make the identification process more accurate. In this paper among the two
technologies that give more accurate results are analyzed. In the proposed method Gray
Level Run Length Matrix (GLRLM) is used for extracting the features. Region-based
segmentation technique is used to segment the affected parts.3D-CNN is used for the
staging and classification of Liver 3D-CT and 3D-US Images separately. The results
are then analyzed and found that 3D-CT gives higher accuracy than 3D-US. The
proposed model is implemented using TensorFlow and Keras of python.
Keywords: Liver 3D CT Image, Liver 3D US Image, GLRLM, SVM.
Cite this Article: A. Bathsheba Parimala and R. S. Shanmugasundaram, An Enhanced
Liver stages classification in 3D-CT and 3D-US Images using GLRLM and 3D-CNN,
International Journal of Electrical Engineering and Technology (IJEET), 12(2), 2021,
pp.171-184.
http://www.iaeme.com/IJEET/issues.asp?JType=IJEET&VType=12&IType=2
1. INTRODUCTION
The liver is a vital organ found on the right side of the abdomen. The liver, which weighs about
3 pounds, is reddish-brown and feels rubbery towards touch. Usually the liver cannot feelsince
it is shelteredusing the rib cage. Chronic liver disease (CLD) is animportantreason of death at
evolved countries.The initial stage is chronic hepatitis. It is caused by the hepatitis B virus. It
causes serious inflammation in the liver. It affects the people mostly from 20 to 60 age group.
The severity of this stage leads to compensated cirrhosis. In this stage, cells degenerate and
inflammation. The fibrous thickening also happened due to viral hepatitis and alcoholism. This
stage is mostly asymptomatic so it remains unsusceptible till the later stage. The later stage is
called decompensated cirrhosis, which has severe symptoms. The symptoms include Ascites,
Jaundice, gastrointestinal bleeding, thrombocytopenia, and hepatic encephalopathy. This leads
to death or hepatocellular carcinoma. The process of diagnosis of CHD is done by clinical data,
invasive method, and noninvasive method. The invasive method is preferred for more accurate
classification. This method includes surgery and biopsy. Though the invasive method gives the

An Enhanced Liver stages classification in 3D-CT and 3D-US Images using GLRLM and 3D-CNN
best result, it involves more complications for some patients. The complication includes
bleeding after biopsy, post-operative problem, pneumothorax, and puncture of the biliary tree.In
the noninvasive methods, images generated by ultrasound, CT, and MRI are used. Medical
image processing plays a major role in the noninvasive method.
Many studies solve this issue, using objective features depending upon US and CT images
with classification procedures to the study of CLD. The general features are defined in terms of
first-order statistic, parallel event matrix, bandwidth change, attenuation, backscattering
parameters, coefficients. SVM and RF classifiers are utilized in the binary classification with
basic CNN neural networks, described on a local basis, integrated into ultrasound and CT image
to presentlocal evidence for adisease
Figure 1 3D Ultra Sound Liver and CT Liver Image
The proposed strategies incorporate division, and Region of Interest depend element
extraction, division, using Gray Level Co-occurrence Matrix (GLRLM) and CNN classifier
calculation, then we can investigate the variation from the norm and help the indicative
apparatus strong to recognize the fluctuation in liver tumor injuries. DICOM formatare used to
represents the medical images. They are highly informative. Along with image data it stores
lots of key patient information, such as, patient’s name, age, sex, doctor’s name etc. When need
to install pydicom python package, using ‘pip install pydicom’. It is better development to the
standard Python prompt, then it ties particularly Matplotlib. IPython begins directly from shell,
or JupyterNotebook.WhenIPython begins, require to attach with graphical user interface event
loop. It specifies IPython to show the layers. To join graphical user interface loop, implement
%matplotlib magic in IPython instigate. IPython's record on graphical user interface event loops
has more information about what it does. It turns in inline plot, where plot graphics can show
the notebook. It contains significant indications of interaction. To in-line plot, instructions in
the cells underneath the output of cell do not influence the plot i.e. change the coloring map
does not feasible from cells under the cell that makes the plot. Nevertheless, to different back-
ends, like tensorflow that open an isolate window, the cells that make up this plot alter this plot
- which has been the direct entity on memory. NumPy has been common array-processing
package. It presents higher-efficiency multi-dimensional array object, then tools to task along
such arrays. This is an essential package of scientific computing and Python. It has several
features. In addition to their clear scientific usage, NumPy could be utilized into a proficient
generic data multiple-dimensional container. Arbitrary data-types could determine utilizing
Numpy that permits NumPy to flawlessly and rapidly incorporate to different types of
databases. SciPy represents open-source software of maths, science, engineering. The library
of SciPy is depending upon NumPy that presents appropriate as well as rapid N-dimensional
array exploitation.
2. RELATED WORKS
This manuscript focuses on the analysis of features removed by applying wave packet
transformation for B-mode ultrasound liver imaging. [1].

This paper summarizes the innovative system for non-invasive, cuff-less, continuous
measurement of blood pressure assessment within patients through NAFLD (non-alcoholic
fatty liver disease) [2].
It defines the unhealthy conditions that are noticed, the effects of three hierarchical
anomalies: 1) chronic hepatitis; 2) compensatory cirrhosis and 3) degenerative cirrhosis. The
characteristics as well as classifiers are applied (Bayes, Parzen, support vector machine and k
nearest neighbor). [3].
This paper examines the diagnostic value of such systems on patients through suspected but
without a definitive diagnosis of cirrhosis. Together transient elastography (TE) and left lobe
liver surface ultrasound (LLS) can detect cirrhosis non-invasively. [4]
This paper discusses the classification of chronic liver disease, a semi-automatic process, to
stabilize that disease depends on liver ultrasound images, medical and laboratory data, using
two classifiers at the center of the algorithm (SVM-Support Vector Machine) (KNN -K-Nearest
neighbor). Classifiers are trained with a multimodal feature test. [5]
The Bayesian classifier is awell-organizedsystem utilized to recover the segmentation of
medical images. To remove background noise from images, the images need to be pre-
processed. After the preprocessing scheme, a Bayesian classifier is employed to classify the
particles on image. [6].
The classification approach is performed using the classifier such as Neural Network. The
classification strategy for liver disease based on the USG of the liver has been established. The
input image undertakes to preprocess that includes Contrast Enhancement, Discrete wavelet
transform, and Segmentation using KNNclassifier [7]
This paper revised various feature extraction techniques. This method employs the Artificial
Neural Network (ANN) forremoving the feature on multiple layers. [8]
To progress for the Image Segmentation process the paper addresses some of the most vital
techniques and represents a survey on them.[9]
The paper presents the current dissimilar segmentation scheme depends on medical images.
This manuscript focuses the work of dissimilar segmentation and classification systems
suggested for the diagnosisof diverse liver diseases [10]. In this paper, automated liver
segmentation by CT imaging remains an open issue as several weaknesses and disadvantages
of the suggested system mayremain be solved. [11]
To evolve an automatic liver segmentation methoddepends on basic knowledge of image,
like location and shape of liver. [12]
The K * methodmay be utilized in analytical tools or instruments to quickly diagnose
specific liver disorders [13]
This paper illustrates the deep learning technique through a convolutional neural network
(CNN) for differentiating liver masses in computed tomography (CT) improvedby dynamic
contrast agent. [14]
This paper identifies the first part as an adaptively weighted cross-entropy loss, which pays
more attention to misclassified pixels. The next part is an edge-preserved smoothness loss,
which assurances neighboring pixels with the same label have related outputs, while
neighboring pixels with different labels have dissimilar outputs. The third part of the loss is a
shape constraint used to model high-level structure differences. The paper shows the data
augmentation both in the training stage and the test stage [15]
This paper illustrates the neural network integrated to categorize liver tumors ashepatoma
and hemageoma. It is executed through modified probability neural network (PNN) in
combination with feature descriptions that are produced by fractal feature as well as gray-level

co-occurrence matrix. A suggested method was assessed via 30 liver cases and proved to be
competent and efficient. [16]
This paper compares the model, with pre-trained model, multi-layered and highly flexible
architectureUsinghe batch normalization to normalize the input layer by adjusting the
activations and to reduce overfitting; we have put the batch normalization before the input after
every layer and dropouts among fully-connected layers.The performance of classifiers is
evaluated by two widely-used measures Matthews Correlation Coefficient (MCC), F1 – Score
(F1 Score is employed for measuring a test’s accuracy). [17]
This paper presents current various methods ofsegmentation based on medical liver images.
And also, this paperfocuses on the work of various segmentation and classificationmethods that
has been proposed to diagnosis many liver diseases [18]
3. PROPOSED WORK
3.1. Data set
The CT Liver image series has been acquired from the Diagnostic Center. The ultrasound image
of the liver is collected in the liver dataset source file.At dataset collection in 1120 CT (3D)
images were setfromprocess of segmentation. The gathered images denotes DICOM format, in
which every image denotes a patient. The gathered data is separated into two sets, 110 normal
CT images and 1000 infected CT images.
Figure 2 Sample Dataset

3.2 Model Architecture
Figure 3 Flow Work
3.3Image Preprocessing
A 3 × 3 × 3 median filter is used to smooth the images, demonstrated at Figure 3. The major
cause for applying median filtering is preprocessing phase of this mechanism is the median
filtering preserve the edge information inside the image in which media filters as well as
Gaussian filters tends for blurring the edges of image. This is the Median filter do not generate
idealistic innovative pixel values in case of filter window is placed above an edge.
The result of this system is the mean value or the mean value of pixel in which entire
intensity distribution values are found, outcoming at innovative image that equivalent the
existing mean values. The medium filter consists ofcapabilityfordecreasing noise on digital
images.
Figure 4 Noisy Image
Input Data (3D-US
and 3D-CT
Preprocessing
(Median Filter)
Segmentation
(Region Based)
Feature
Extraction
(GLRLM)
Classification
using 3D-CNN
Segmented
Abnormal
Image
Abnormal
Normal
Classify Stages
using 3D-CNN
Stages
Prediction
Stage 1
Stage 2
Stage 3
Stage 4
Classification
Phase-I
Classification
Phase-II

Figure 5 De-Noise Image
3.4 3D US and CT Liver Image Classification
In the Classification phase the segmentation is first executed using region-based method. Then
GLRLM [19] is implemented to extract the features. All the process is done for both the US
and CT images.
3.4.1 Region based segmentation
For the segmentation process, the 3D region growing method (3D painting algorithm) is used.
It fills entire voxels that consists of CT values minimum to provide threshold value to similar
color begins as provided starting point. This filled voxels equivalent with connected affected
region. This algorithm is executed in sequence, line-by-line mode beside axes on 3D space for
diminishing computation time. Begins in starting point which should be an internal Region of
Interest (ROI)[20] point, it investigates the adjacent onefordecidingthe CT values less than the
provided threshold value.
Figure 6 Segmented Image
3.4.2 Feature Extraction
GLRLM (Gray Level Run Length Matrix)
This strategy is an optimal method of extracting high request measurable surface highlights.
Consider be the dark level quantity, long run, number or pixels at the picture. The GLRLM is

2D grid of (G*R) components, where every component provides all out count of run events
contains length j of dark level I, at provided guidance. GLRLM assists with trimming locale of
intrigue physically from a picture as well as its processes seven surface boundaries like short
run accentuation, and since quite a while ago run emphasis, dim level non-consistency, run rate,
run length non-consistency (RLN), less dark level run accentuation, maximal dim point run
accentuation.
Table 1 GLRLM Features using pyradiomix
GLRLM FEATURES VALUES
GrayLevelNon-Uniformity 175.6351923150419
GrayLevelNon-Uniformity
Normalized
0.04514123814981055
GrayLevelVariance 39.118151021979244
Maximum GrayLevelRunEmphasis 281.066493908972
ExtendedRunEmphasis 1.2268440382584342
ExtendedRun Maximum GrayLevelEmphas
is
341.2865790983503
ExtendedRun Minimum GrayLevelEmphas
is
0.010601170478748765
Minimum GrayLevelRunEmphasis 0.008600397891661503
RunEntropy 4.915038003159503
Run LengthNon-Uniformity 3500.0432315746298
RunLengthNon-Uniformity
Normalized
0.8950494659480998
RunPercentage 0.9404064632491029
RunVariance 0.08479457789590625
ShortRunEmphasis 0.9559391731405504
ShortRun Maximum GrayLevel
Emphasis
268.9741798411307
ShortRun Minimum GrayLevel
Emphasis
0.008229766244155428
Classification using 3D-CNN
Two stages of classification is carried out. First stage predicts the image as normal or abnormal.
The second stage gets the abnormal image as input and predicts the four stages of Image (Fatty
Liver, compensated Cirrhosis, Decompensate Cirrhosis, Hepatocellular Carcinoma (HCC)
3D CNN consists of convolution layers in which the first convolution layer contains 8
Filters size 7X7X7. Second convolution layer has 16 filters size 5X5X5 and the third convolution
layer has 32 filters size 3X3X3. Next to the convolution layers, pooling layer is placed that is
in charge todiminish the convolved features spatial size. In the proposed model Maxpool is
used. The kernel of 3X3X3is used to Maxpoolwhere maximal value of the part of the image
enclosedwith kernel is got.RelU activation function is performed in which the function outputs
the given input straightly if it is positive (> 0); or else, it will give an output of 0. A last layer
is a completelylinked layer that is used for classifying the stages of liver abnormalities. The
classifier is required after feature removal to locate the equivalent label of all test image. After
the layer is completelylinked, the Softmax activation is applied to obtain the stage
classification.It converts the value in the fully connected layer in to a probability distribution.

Architecture of the CNN for liver classification.
The first convolutional layer performs 7X7X7 convolutions with a 5×5×5 stride and with valid
padding; followed by a ReLU layer, pooling layer size 3×3×3. The second convolution layer
performs 5X5X5 convolutions through a 4×4×4 stride and with valid padding; followed by a
ReLU layer, pooling layer size 3×3×3. The next convolution layer performs 3×3X3 convolutions
with a 2×2×2 stride and with valid padding; followed by a ReLU layer. The next convolutional
layer performs 2×2×2convolutions with a 2×2×2 stride and with valid padding; followed by a
ReLU layer. The next convolutional layer performs 3×3×3 convolutions with a 1×1×1 stride
and with valid padding; followed by a ReLU layer, pooling layer of size 2×2×2. The next layer
is a 0.4% dropout layer in this work dropout is used to regularize dense layers.Two phases of
classification is carried out. In the first phase of classification the segmented image is classified
in to normal or abnormal. Then the segmented abnormal image is again classified in the second
phase of classification.The output Layer is formed by 4 output nodes for classifying four
classes.
Figure 7 Chronic Hepatitis
Figure 8 Compensated cirrhosis

Figure 9 Decompensated cirrhosis
4. RESULTS AND DISCUSSIONS
To break down the proposed presentation framework for detecting the tumors, images acquired
utilizing the proposed procedure have been contrasted and relating ground truth images.
Various measures are usually used to assess the presentation of the technique. The
characterization precision (AC), Correlation Coefficient (CC) have been determined from
disarray grid. The disarray framework depicts actual and predicted classes of the technique.
Confusion Matrix for US Image
Predicted
Result
Negative
Positive
True Values
Positive Negative
653
TP
32
FP
38
FN
387
TN
Figure 10 Confusion Matrix for CT
The model is evaluated using the confusion matrix which is used to evaluate precision,
recall, F1score and accuracy. Based on the results, two confusion matrices have been formed
individually for Ultrasound image and Computerized Tomographical image. Four stages have
been predicted and the overall accuracy is calculated with the help of confusion matrix.
Precision Value (Pr) = No.of correct +vePrediction(TP) /No.ofcorrect +ve prediction +
False +ve prediction (FP)
Recall Value (Rc)= No. of correct +ve Prediction /No. of correct +ve Prediction + False -ve
Prediction (FP)
F1Score = 2 * Pr * Rc/Pr + Rc

Accuracy = No. of correct +vePrediction+No.of correct -ve Prediction / (No.of correct +ve
prediction + No.of correct -ve Prediction + No.of False +ve prediction + No.ofFalse -ve
prediction)
Prediction chart for Ultrasound Image
Table 2 Prediction table for Ultrasound Images
Stages
No. of
samples
Input
Correct
Predic
tion
Wrong
Predic
tion
Accu
racy
Fatty Liver 200 184 16 0.92-
Decompensated
Cirrhosis
160 148 12 0.93
Compensated
Cirrhosis
290 270 20 0.93
Hepatocellular
Carcinoma
350 332 18 0.95
Normal 110 102 8 0.93
Total 1110 1036 74
0.93
Average accuracy
Confusion Matrix for CT
Predicted
Result
Negative
Positive
True Values
Positive Negative
683
TP
22
FP
14
FN
391
TN
Figure 11 Confusion Matrix for CT
Prediction chart for Ultrasound Image
Table 3 Prediction table for CTImages
Stages No. of
samples
Input
Correct
Predic
tion
Wrong
Predic
tion
Accu
racy
Fatty Liver 200 192 7 0.96
Dcompensated
Cirrhosis
160 153 6 0.947
Compensated
Cirrhosis
290 281 8 0.968
Hepatocellular
Carcinoma
350 342 7 0.977
Normal 110 106 4 0.963
Total 1110 1074 36
0.965
Average accuracy

In this section, the effectiveness and execution of the liver disorder technique are depicted
dependent on the pre-processing, feature extraction, include selection, and characterization. The
exhibition of the proposed try was tried in PYTHON by using medical images. The medical
images joined standard CT and Ultrasound images of the liver and different unhealthy liver
images are used.
Table 3 Accuracy Comparison Table
Technology
Recall
(Sensitiv
ity)
Precision
F1
Score
Accuracy
3D CNN
For CT
Image
0.98 0.96 0.97 0.96
3D-CNN for
US Image
0.95 0.94 0.94 0.94
Table 3 above shows the exactness displayed by Ultrasound Image diverse characterization
strategies and CNN Model. It is seen that precision of proposed model gave 94% for US images,
96% for CT images
Figure 12 Metrics values Comparison Chart
The examination shows the achievability of utilizing profound learning for the liver. In spite
of the fact that we utilized a little dataset containing images from certain patients, this
information was adequate to build up a well performing classifier with move learning. In
addition, the CNN-based methodology accomplished altogether preferred outcomes over the
GLRLM-basedmethodology. CNN features were helpful and empowered productive preparing
of the grouping and relapse models. Great execution of the CNN-based methodology was
expected. In our examination, we didn't prepare the system without any preparation; rather the
pre-prepared CNN was utilized for include extraction.
0.92
0.93
0.94
0.95
0.96
0.97
0.98
0.99
(Sensitivity)
Recall Precision F1 Score
Comparison Chart
3D CNN For CT Image 3D-CNN for US Image

Figure 13 Accuracy chart
5. CONCLUSION
3D-US and CT images are used as higher dimensional images give more information than lower
dimensional images. 3D-CNN is used to classify liver disorders. The comparison of the
performance of 3D-CNN in US and CT images are carried out in this paper. The region growing
segmentation along with the features extracted through GLRLM Combined with 3D-CNN helps
to improve the prediction accuracy. The result and analysis show that 3D-CNN for CT images
give more accuracy than US images. We have used the python for coding and we used Keras
to build our CNN. For future work we need more data for getting higher accuracy result.
REFERENCES
[1] Jiuqing Wan and Sirui Zhou, Features Extraction Based on Wavelet Packet Transform for B-
mode Ultrasound Liver Images. International Congress on Image and Signal Processing 2003.
[2] Trine M. Seeberg - Member IEEE, James G. Orr, Hanne A novel method for continuous, non-
invasive, cuff-less measurement of blood pressure: evaluation in patients with non-alcoholic
fatty liver disease
[3] Ricardo T. Ribeiro, RuiTatoMarinho, and J. Miguel Sanches, Classification and Staging of
Chronic Liver Disease from Multimodal Data. IEEE Transactions on Biomedical Engineering.
May 2013.
[4] Annalisa Berzigotti and Juan G. Abraldes, Ultrasonographic evaluation of liver surface and
transient elastogr aphy in clinically doubtful cirrhosis. European Association for the
study of Liver. Journal of Hepatology. 2010.
[5] Ricardo Riberio, RUI TatoMarinho, Jose velosa,FernandoRamalho, J. Miguel Sanches and
Jasijit S. Suri, The Usefulness of Ultrasound in the classification of Chronic Liver Disease. IEEE
Xplore. 2011
[6] M.C.Jobin Christ, K.Sasikumar, and R.M.S.Parwathy, Application of Bayesian Method in
Medical Image Segmentation. TECHNIA International Journal of Computing Science and
Communication Technologies, July 2009
0.93
0.935
0.94
0.945
0.95
0.955
0.96
0.965
3D CNN For CT Image 3D-CNN for US Image
Accuracy

[7] Vishakha V. Hambire1, Dr. S. R. Ganorkar2 “Classification of Liver Disease Based on US
Images”International Research Journal of Engineering and Technology (IRJET) Volume: 02
Issue: 04 July-2015.
[8] RajkumarGoel1,VineetKumar,SaurabhSrivastava , A. K. Sinha, A Review of Feature Extraction
Techniques for Image Analysis.International Conference on Advances in Computational
Techniques and Research Practices Noida Institute of Engineering & Technology, Greater
NoidaVol. 6, Special Issue 2, February 2017
[9] Survey,RahulBasak, Surya Chakraborty, Aditya Kumar Mondal, SatarupaBagchiBiswasImage
Segmentation Techniques: A International Research Journal of Engineering and Technology
(IRJET)Volume: 05 Issue: 04 | Apr-2018.
[10] Medical Image Segmentation for LiverDiseases: A Survey,Tarek M.
Hassan,MohammedElmogy,ElsayedSallam.International Journal of Computer Applications
(0975 – 8887)Volume 118 – No.19, May 2015
[11] Ahmed M. Mharib · Abdul RahmanRamli ·SyamsiahMashohor · RoziBintiMahmoodSurvey on
liver CT image segmentation methods,.Published online: 26 April 2011Springer
Science+Business Media B.V. 2011
[12] Mohamed Ali Mahjoub,Automatic liver segmentation method in CT images,OussemaZayane
,BesmaJouiniCanadian Journal on Image Processing & Computer Vision Vol. 2, No. 8,
December 2011
[13] Shapla Rani Ghosh and SajjadWaheed,J. Sci. Technol. Analysis of classification algorithms for
liver disease diagnosis, Environ. Inform. 05(01): 361-370 | Ghosh and Waheed (2017).
[14] Deep Learning with Convolutional Neural Network for Differentiation of Liver Masses at
Dynamic Contrast-enhanced CT: A Preliminary Study, KoichiroYasaka, Hiroyuki Akai,Osamu
Abe, Shigeru Kiryu,,Radiology: Volume 286: Number 3March 2018.
[15] Man Tan, Xiongwei Mao, Fa Wu, and Dexing Kong, Liver segmentation using 3D CNNs with
high level Shape constraints
[16] E-Liang Chen, Pau-Choo Chung, Ching-Liang Chen, Hong-Ming Tsai, and Chein-
IChang,anAutomatic Diagnostic System for CT LiverImageClassification, IEEE Transactions
on biomedical Engineering, vol. 45, no. 6, june 1998
[17] Classification of Liver lesions by Using CNN in CT Images May 20, 2020
[18] A. BathshebaParimala, R. S. Shanmugasundaram, Assessment on Liver Disease
Classificationusing Medical Image Processing,International Journal of Recent Technology and
Engineering (IJRTE)
[19] Jefferson, Shanmugasunadaram,Assessment on Brain Tumor Detection Techniques in
HyperIntense MR Images,International Journal of Recent Technology and Engineering
(IJRTE).
[20] A. Bathsheba Parimala, R. S. Shanmugasundaram, Liver Staging And Classification: Using
Classifier, Image Test And Clinical Test,International Journal of Computer Engineering and
Applications,Volume- XIV, Issue - Special Issue, June 2020, www.ijcea.com ISSN 2321-3469

AUTHORS PROFILE
A.Bathshebaparimala, has received M.Sc.
He is currently a research scholar working toward the PhD degree in the
Department of Computer Science, VMKV Arts and Science College, Salem. Her
research interests are in Medical image processing, Medical image analysis and Image
Classification.
Dr.R.S.Shanmugasundaram, has received the BE degree in Electronics and
Communication Engineering from the university of Madras of India in 1996, ME
degree from Bharathidasan University of India in 2001 and Ph.D degree from
Anna University of India in 2015. His research interests are in medical image
processing, deformable models and segmentation. He is a life member of ACS and ISTE.
Currently he is working as Deputy Director (Academics) of Vinayaka Mission’s Research
Foundation (Deemed to be University), Salem, Tamil Nadu, India

An enhanced liver stages classification in 3 d ct and 3d-us images using glrlm and 3d-cnn

More Related Content

What's hot (18)

Similar to An enhanced liver stages classification in 3 d ct and 3d-us images using glrlm and 3d-cnn (20)

More from Bathshebaparimala (20)

Recently uploaded (20)

An enhanced liver stages classification in 3 d ct and 3d-us images using glrlm and 3d-cnn