Liver Tumor Segmentation Using SVM Framework and Pathology Characterization Using Content-Based Image Retrieval

This chapter presents the semiautomated segmentation of liver tumor from computed tomography (CT) scans under a hybrid support vector machine (SVM) framework and a content-based image retrieval prototype system based on multiphase CT images to support the decision-making for liver tumor characterization. It presents a three-stage, hybrid support vector machine (HSVM)-based approach for liver tumor segmentation. In this method, HSVM is a seamless and natural connection of one-class support vector machine (OSVM) and binary support vector machine (BSVM) by a boosting tool. The chapter introduces a content-based image retrieval (CBIR) prototype system based on multiphase CT images to help radiologists in characterizing focal liver tumors. With further development and validation, these methods have the potential of being adopted as image analysis tools to assist liver tumor volumetry and characterization for cancer diagnosis, treatment planning, and assessment of therapy response.