Eigenmotion-Based Detection of Intestinal Contractions Laura Igual 1 , Santi Segu´ ı 1 , Jordi Vitri` a 1 , Fernando Azpiroz 2 , and Petia Radeva 1 1 Computer Vision Center and Universidad Aut´onoma de Barcelona, Bellaterra, Spain 2 Hospital de Vall d’Hebron, Barcelona, Spain Abstract. Intestinal contractions are one of the main features for ana- lyzing intestinal motility and detecting different gastrointestinal patholo- gies. In this paper we propose Eigenmotion-based Contraction Detection (ECD), a novel approach for automatic annotation of intestinal con- tractions of video capsule endoscopy. Our approach extracts the main motion information of a set of contraction sequences in form of eigen- motions using Principal Component Analysis. Then, it uses a selection of them to represent the high dimension motion data. Finally, this con- traction characterization is used to classify the contraction sequences by means of machine learning techniques. The experimental results show that motion information is useful in the contraction detection. More- over, the proposed automatic method is essential to speed up the costly examination of the video capsule endoscopy. 1 Introduction The analysis of the small bowel contractions has been proved to be a mean- ingful method for diagnosing several intestinal dysfunctions [1], [2]. Currently, the most extended diagnosis test for motility disorders is intestinal manometry [3], which measures variations of pressure. This technique is highly invasive and disagreeable and requires hospitalization of the patient. A recent and very promising alternative acquisition method is the Wireless Capsule Video Endoscopy (WCVE) [4] which consists of a capsule with a camera, a battery, and a set of led lamps for illumination. The capsule is swallowed by the patient and emits a radio frequency signal stored in an external device. The result is a video which records the ”trip” of the capsule along the intestinal tract with a rate of two frames per second. This novel technique is non-invasive compared with the manometry and there is no need of hospitalization of the patient. The result- ing images provide a view of the inner gut in which the gut wall and lumen are visualized (Fig. 1 (right)). The visual paradigm of intestine muscle contractions in WCVE is defined as a close-open movement of the lumen. Fig. 1 (left) shows three examples of intestinal contraction sequences. They are defined by 9 frames, since phasic contraction takes on average 4.5 seconds. The non-occlusive contractions (lumen is not completely closed) are hard to detect by classical manometry, since the intestinal walls are not accomplishing enough amount of pressure. However, W.G. Kropatsch, M. Kampel, and A. Hanbury (Eds.): CAIP 2007, LNCS 4673, pp. 293–300, 2007. c  Springer-Verlag Berlin Heidelberg 2007