Identification of Intestinal Motility Events of Capsule Endoscopy Video Analysis

Purpose: To develop a system for assisting the analysis of capsuleendoscopy (CE) video data and identifying sequences of frames related to small intestine motility. Material and Methods: Six videos were analyzed and labelled manually by an expert, indicating the events of contractions. For addressing the imbalanced recognition task of small intestinal contractions we employed an efficient two-level video analysis system. At the first level of the system, each video was processed resulting in a number of possible sequences of contractions. In the second operating part of the system, the final recognition of contractions sequences was carried out by means of a SVM classification algorithm. To encode the patterns of intestinal motility a panel of textural and morphological features of the intestine lumer were extracted. Results: The system exhibited an overall sensitivity of 73.53% in correct detecting events of contractions. The false alarm ratio (false positives over the true positives) was of the order of 59.92%.Conclusion: these results serve as a first step for developing assisting tools for computer based CE video analysis, reducing drastically the physician’s time spent in image evaluation and enhancing the diagnostic potential of CE examination.