Abstract
We have developed a useful system to automatically segment colonic wall based on anatomical knowledge form a series of abdomen computer tomography volume dataset. It employs distance from boundary, potential field and volumetric rendering to identify the true polyps inside human colon. The proposed system also provides an interactive user interface for navigation with virtual rendering and is applied to the sub-volume of polyp candidates with 3-D shape, gray level and curvature. It can efficiently detect polyps with two-dimensional viewing images parallel to axial, sagittal and coronal in an interactive manner. With reconstruction of CT volume data, a virtual colonography system can be used to simulate the examination of fiber optical colonography. It can help the radiologists classify the polyps in clinical diagnosis.