Segmentation of liver tumor from CT images

Abstract

This final year project aims to study characteristics of cancerous and healthy tissues in liver, which are later used by two algorithms, Support Vector Machine and Graph-cut, to do the liver tumor segmentation. Results from two algorithms are compared to evaluate on their efficiency. Besides, anisotropic diffusion filter concept is also introduced as a pre-processing step, so that the final segmentation from both algorithms can be enhanced. Finally, a new semi-automatic method is proposed to improve liver tumor segmentation by finding out the ribcage curves on every CT slice, using the rib bone traces. One can use the rib curve as a post-processing step to fine tune the segmentation.