Comparative study of contour detection methods for intestinal sessile polyps

The past decade represented a step forward in a new direction in the exploration of the small bowel, with the introduction of the wireless capsule endoscopy (WCE). Lesions like intestinal polyps became easier to diagnose in early stages, reducing thus the risk of small bowel cancer. Since the curved shape of polyps represents the most important feature in their automatic identification, proper edge detection techniques are necessary for emphasizing their contour. This paper presents a comparative study of three edge detection methods applied on a set of original and noise processed WCE images containing sessile polyps, based on correct edge localization and a proper contour detection. The oriented Gabor phase congruency method presented better results compared to Sobel and Canny operators, both for original and processed images, recommending it for further use in computer-aided diagnostic systems.