An abnormality based WCE video segmentation strategy

Wireless Capsule Endoscopy (WCE) is a state-of-the-art technology, which allows complete exploration of the small intestine. Despite clinical findings that WCE videos are promising, there still exist several problems. The most crucial problem is that it is a highly time-consuming task for physicians to inspect the entire video. So it is necessary to investigate CAD based automatic diagnosis system to reduce the burden of physicians. In this paper, we propose a novel scheme to catalogue the WCE video clips with respect to abnormalities instead of organs. The aim of the proposed scheme is to provide an alternative option to doctors in hope to increase the accuracy of the diagnosis as well as reduce the inspection time. The novel method is based on the adaptive non-parametric key-point detection using multi-feature extraction and fusion. Actual clinical patient videos including both normal and abnormal findings are used to evaluate the performance of the proposed method. The experimental results demonstrate that the proposed approach leads to efficient segmentation for WCE video clips without losing critical information of the original video record.