Gland segmentation guided by glandular structures: A level set framework with two levels

Abstract

Pathologic diagnosis is the gold standard of clinical diagnosis. The identification and segmentation of histological structures are the prerequisites to disease diagnosis. In clinic, doctors often suffer from time consuming and the disagreements from different doctors about observation results. Hence, an automatic precise segmentation method is important for auxiliary diagnosis. We propose a level set framework using 0, k level set representing the boundary of lumen regions and epithelial layers for gland segmentation. The validation has been performed on clinical data of West China Hospital, Sichuan University. The experiment results show that our method has a better performance and is robust to the shape variety of endometrial glands.