Fast Graph Cuts Based Liver and Tumor Segmentation on Olumetric CT Images

Abstract

Accurate segmentation of liver and tumors from abdominal computer tomography (CT) scans is critical for computer-assisted diagnosis and therapy. In this paper, we proposed a graph cuts based segmentation approach for fast automatic liver segmentation and semi-automatic liver tumor segmentation from CT images. Only one seed point for each tumor was selected manually. Liver/tumor volumes of interest (VOI) were extracted automatically and supervoxels were employed to reduce the computational cost. Intensity and spatial information were incorporated into the graph cuts algorithm. Experimental results demonstrate that the proposed method can detect the liver and tumor accurately with significant reduction of processing time.