Abdominal Imaging. Computational and Clinical Applications

Abstract

Low contrast between tumor and healthy liver tissue is one of the significant and challenging features among others in the automated tumor delineation process. In this paper we propose kernel based clustering algorithms that incorporate Tsallis entropy to resolve long range interactions between tumor and healthy tissue intensities. This paper reports the algorithm and its encouraging results of evaluation with MICCAI liver Tumor Segmentation Challenge 08 (LTS08) dataset. Work in progress involves incorporating additional features and expert knowledge into clustering algorithm to improve the accuracy. © 2012 Springer-Verlag.