Segmentation of retroperitoneal tumors using fast continuous max-flow algorithm

Abstract

In this paper a new algorithm for the segmentation of retroperitoneal tumors is presented. In this paper an accumulated gradient distance is proposed as a new regional term in the continuous max-flow algorithm. In this preliminary study, three CT images have been segmented and compared with manual segmentations. DICE, Jaccard, Sensitivity and Postive Predictive Value (PPV) indexes have been computed and their values show promising results. © Springer International Publishing Switzerland 2014.