Liver of CT Slice Localization and Segmentation Based on GPV-levelset

Abstract

Motivated by the goal of improving segmentation of challenging liver cases, which always contain weak boundary with neighboring organs and the presence of intrahepatic tumors as well as highly varied appearance between different subjects, a multi-level detection segmentation framework is presented. The Gaussian pseudo-variance based on local patch is utilized to improve the level set segmentation technique, which named as Gaussian pseudo-variance level set (GPV-levelset). Our method considers the liver scale, the spatial texture, and the changes of the energy functional over iterations. Therefore, the image with uneven gray scale and noise images can be processed effectively. We demonstrate the capabilities of the method in the domain of medical imaging for segmenting two public databases Sliver07 and 3Dircadb.