Multiple classification method for analysis of liver lesion with focal liver segmentation techniques for CT image

Abstract

Liver is the biggest organ in the human body, which has basic functionalities like storage of glucose, producing bile juice. In this paper the analysis of focal liver lesions of CT image using multiple classifier methods is addressed. The analysis scheme includes pre-processing, segmentation, feature extraction and classification. Pre-processing of the image is through anisotropic diffusion to reduce noise and inhomogenities in the image. Segmentation with seeded region growing is followed by binary masking to extract liver from CT image. Five features that are based on texture properties such as mean, standard deviation, entropy, root mean square (rms) value and energy are extracted. Classification using Neural Networks, K-Nearest Neighbour (KNN) and Support Vector Machine (SVM) to differentiate the normal image and abnormal imageis performed. In Neural Networks receiver operating curve and confusion matrix are used to evaluate the performance of the system for different number of hidden layers. In KNN and SVM the performance of the system is evaluated based on sensitivity, specificity, accuracy and predictive positive value. Finally, comparison of multiple classification methods is done to evaluate the best classification approach for focal liver lesions of CT image. © 2013 Springer.