Extraction and Classification of Liver Abnormality Based on Neutrosophic and SVM Classifier

Liver is the important organ and common site for a variety of cancer diseases. The most important steps in treatment planning and evaluation of liver cancer are to identify the presence of liver cancer and to determine the various stages of liver cancer. This paper proposes an automatic method to segment the liver from abdominal computer tomography imaging and classify the liver as normal or abnormal liver. The aim of this work is to develop computer-aided liver analysis to segment the liver and classify the liver, thereby helping the physician for treatment planning and surgery. The method uses median filter for preprocessing and neutrosophic (NS) domain with FCM thresholding for segmenting the liver. In post processing, morphological operation is done to obtain liver contour. Features are extracted from the segmented liver using gray-level co-occurrence matrix (GLCM). These feature vectors are given as input to train the support vector machine (SVM) classifier, to classify healthy or unhealthy liver. The classifier performances are assessed and analyzed using various quality metrics like accuracy, sensitivity, specificity and misclassification rate.