A Novel Computer Aided Detection (CADe) Scheme for Colonic Polyps Based on the Structure Decomposition

Abstract

Accurately detecting small polyps (ranged from 5~8mm) on the colon wall is of great significance for early diagnosis colorectal cancers. However, colon usually consists of the mucosa layers which result in partial volume effect (PVE) on the colon wall. Consequently, the task of computer aided detection (CADe) of polyps turns into too complicated to be reached by simply following solo philosophy. In order to achieve the mission of small polyps’ detection, we propose a novel global structure decomposition approach in this paper. That is, the complex colon was separated into much uniform broken parts by means of analysis on second order derivatives of the volume image. Experimentally, we chose 60 patient cases from dataset provided by Wisconsin, and in which we focus on the polyps whose size range from 5~8mm to validate the presented new approach. Compared with previously presented in the literature, the experimental results are much more promising with an average sensitivity of 0.984. Meanwhile, the false positive rate dramatically decreased to 2.2 per dataset after false positive reduction.