Polyp detection in wireless capsule endoscopy videos based on image segmentation and geometric feature

Abstract

Wireless Capsule Endoscopy (WCE) is a relatively new technology (FDA approved in 2002) allowing doctors to view most of the small intestine. One of the most important goals of WCE is the early detection of colorectal polyps. In this paper an unsupervised method for the detection of polyps in WCE videos is presented. Our method involves watershed segmentation with a novel initial marker selection method based on Gabor texture features and K-means clustering. Geometric information from the resulting segments is extracted to identify polyp candidates. Initial experiments indicate that the proposed method can detect polyps with 100% sensitivity and over 81% specificity. ©2010 IEEE.