Home // International Journal On Advances in Intelligent Systems, volume 5, numbers 1 and 2, 2012 // View article

Lumen Detection in Endoscopic Images: a Boosting Classification Approach

Authors:
Giovanni Gallo
Alessandro Torrisi

Keywords: Pattern Recognition; Boosting; Wireless Capsule Endoscopy; Video Automatic Annotation; Support Vector Machine.

Abstract:
Intestinal lumen detection in endoscopic images is clinically relevant to assist the medical expert in study- ing intestinal motility. Wireless Capsule Endoscopy (WCE) produces a high number of frames. Automatic classification, indexation and annotation of WCE videos is crucial to a more widespread use of this diagnostic tool. In this paper we propose a novel intestinal lumen detection method based on boosting. In particular, we propose a customized set of Haar- like features combined with a variant of AdaBoost to select discriminative features and to combine them into a cascade of strong classifiers. Experimental results show the efficacy of boosted classifiers to quickly recognize the presence of intestinal lumen frames in a video. To better assess the accuracy of the proposed boosted classifier, we present an experimental comparison with the results obtained with a Support Vector Machine using a linear kernel.

Pages: 127 to 134

Copyright: Copyright (c) to authors, 2012. Used with permission.

Publication date: June 30, 2012

Published in: journal

ISSN: 1942-2679