ABSTRACT

Several characteristics of liver sonography have been used to evaluate the degree of severity of liver disease, including changes of the liver surface, inferior edge, echotexture, echogenicity, and diameters of hepatic and splenic veins. This chapter proposes a computerized tissue classification system to detect the presence of focal liver disease based on texture analysis. A brief description of sonographic appearances of different liver image classes considered. The chapter presents a study on the performance of Computer-Aided Classification (CAC) systems based on single principal component analysis (PCA)-neural network (NN) based multi-class classifier design has been compared with CAC system based on PCA-NN based hierarchical classifier design. The main conclusion of the present research work is that CAC system designs using texture features and texture-ratio features collectively enhance the performance of the system for characterization of focal liver lesions (FLLs).