Intestinal Motility Evaluation by Endoluminal Image Analysis Using Capsule Endoscopy (CE-EIA): a Multi-center Clinical Trial

Conventional intestinal manometry is the current gold standard for the evaluation of intestinal motility1,2,3,11, and identifies patterns of intestinal dysmotility4,8,9. However intestinal manometry involves intestinal intubation with consequent discomfort for the patients. Furthermore, it requires considerable technical expertise and knowledge for interpretation of the data. Hence, to date this method has limited indications and is restricted to very few referral centers around the world6,7,10,13.

Vall d'Hebron Hospital, in collaboration with CVC (Barcelona, Spain) have recently developed a minimally invasive method for evaluation of intestinal motility based on endoluminal image analysis using the endoscopic PillCam capsule. In contrast to manometry, this technique is minimally invasive, the technical aspects are simple, and the analysis is performed fully automated by a computer program. Both the technical procedure of the test and the endoluminal image analysis program has been developed by a multidisciplinary medical-engineering team in the Autonomous University of Barcelona over the past 5 years. The technique has been validated in a group of patients with intestinal dysmotility and healthy subjects, and has demonstrated over 90% sensitivity and specificity.

In brief, the technique works as follows. In each study a series of features are analyzed: contractile patterns (contractions evaluated as a diaphragmatic occlusion of the lumen and by the presence of a radial wrinkle pattern), non contractile patterns (wall and tunnel patterns), luminal content (turbid pattern), endoluminal motion, and capsule displacement.

The program is based on an automated learning method (machine learning technique). Data from patients and healthy subjects are used as a training set. Based on these data, the program develops the function that best discriminates both groups. The performance of the system has been validated using the leave-one-out method that uses all but one as training set and evaluates the left-out example.

This technique needs now to be validated in a large multinational population. Using this expanded pool of studies as a learning set, a robust discrimination algorithm will be developed, that can be made available for widespread diagnostic application. Furthermore, whereas manometry only recognizes neuropathic, myopathic5,12 and obstructive motor patterns, endoluminal image analysis may identify different categories of patients depending on the clinical presentation and the etiologic factors involved.