Computer-Aided Diagnosis of Liver Cirrhosis Based on Multiple Statistical Shape Models

Abstract

In the fields of medical image analysis and computational anatomy, statistical shape models (SSMs) is usually used for organ segmentation; SSMs are statistically constructed from a population of organs. In this paper, we focus on the application of SSMs for the computer-aided diagnosis of cirrhotic livers. Since chronic liver diseases or cirrhosis will cause significant morphological changes on both the liver and spleen, we constructed multiple SSMs (i.e., liver SSM, spleen SSM, and a joint SSM of the liver and spleen) for morphological analysis. Coefficients of SSMs are used as features for the classification of normal and cirrhotic livers. Through this paper, we show that classification accuracy can be significantly improved by effective mode selection, which is based on fisher discriminant analysis, and the use of a non-linear support vector machine. Furthermore, we also construct Computer-aided Diagnosis (CAD) of liver cirrhosis system using SSMs. Keywords-computational anatomy; statistical shape model; liver cirrhosis; computer-aided diagnosis; effective mode selection; fisher discriminant analysis; non-linear SVM