Automatic classification of frames from wireless capsule endoscopy

Abstract

Wireless Capsule Endoscopy (WCE) is a technical break-through that allows to produce a video of the entire intestine without surgery. Interpretation of WCE video is nowadays largely left to the visual inspection of a medical specialist. This tedious and time consuming task could greatly benefit from techniques that automatically classify and exclude from further processing the non-relevant frames in the video. In this dissertation several distinctive methods to tackle the problem of automatic classification of image frames belonging to a WCE video are presented. In particular we focuses two areas: sudden changes discrimination and intestinal motility detection in a WCE video. The achieved high detection accuracy of the proposed systems have provided an indication that such intelligent schemes could be used as a supplementary diagnostic tool in endoscopy.