A Learning Method for Automated Polyp Detection

Adenomatous polyps in the colon have a high probability of developing into subsequent colorectal carcinoma, the second leading cause of cancer deaths in United States. In this paper, we propose a new method for computer-aided diagnosis of polyps. Initial work with shape detection has shown high sensitivity for polyp detection, but at a cost of too many false positive detections. We present a statistical approach that uses support vector machines to distinguish the differentiating characteristics of polyps and healthy tissue, and subsequently uses this information for the classification of the new cases. One of the main contributions of the paper is a new 3-D pattern analysis approach, which combines the information from many random images to generate reliable signatures of the shapes. The proposed system has a good polyp detection rate with the false positive rate reduced by 80 percent compared to previous work.