Comparative Analysis on Performance of Different Classification Models for Accurate Liver Tumor Segmentation and Classification

Abstract

One of the most common cancer disease is the occurrence of tumors in liver organ. Hepatocellular carcinoma is one of the most common type of cancer found in liver organ which causes most mortality in the world. The imaging technique mostly used to detect these type of cancer is Computed Tomography (CT) scans. In this paper we have proposed computer aided diagnostic system for detection of tumors in liver organ and its classification model to grade the tumor stage in the liver organ on the basis of tumor volume. We have made a comparative analysis of the different classification model used for this proposed model. The classification techniques used are SVM, KNN, Random Forest classifier, Naïve Bayes classifier and linear discriminant analysis. A total of 11 patients CT scans are considered for this study and real world datasets are collected from a diagnostic center in Guwahati, Assam. Each patient dataset consists of 800 to 1200 CT images captured at different contrast levels to enhance the quality of the image captured. The analysis of the classification model is done by considering the accuracy of the model and volumetric error overlap(VOE). It has been found from the result analysis that SVM outperforms all other classifiers in terms of accuracy and VOE.