Automated liver and tumor segmentation based on concave and convex points using fuzzy c-means and mean shift clustering

Abstract

Identifying liver and tumour regions from medical images using fully automated computer-aided software is a challenging task for the diagnosis of liver disease. In this paper, a novel method is presented to overcome the problems of liver and tumour segmentation in CT images such as weak boundaries, touching organs, and heterogeneity of the liver. Organ edges were extracted using the Kirsch filter, and the concave and convex points were subsequently calculated. The mean-shift algorithm was employed to make the images uniform along the organ borders. Finally, the FCM approach was used to segment the liver and tumours. The results demonstrated that our complex algorithm achieved an average surface distance (ASD) of 1.1±0.39mm and volume overlap error (VOE) of 1.8±0.34% while segmenting the liver. An ASD of 1.5±0.55mm and VOE of 9.8±3.9% were obtained with respect to tumour segmentation.