Detection of Colonic Landmarks for Supine-Prone Registration

In virtual colonoscopy, a patient is scanned in both supine and prone positions. However, since the shape of the colon changes substantially, it is necessary to find a good registration between the two scans to help the physician with better polyp detection and to reduce the false positives. To obtain a registration, anatomical features are needed. We use anatomical landmarks to have the necessary orientation information, as they do not change despite the change in the position of the person. The taeniae coli and the flexures form important anatomical landmarks which can be used to provide the orientation information. Taeniae coli can be viewed as ridge breakers for the haustral folds. The haustral folds are detected by using the heat diffusion and fuzzy C-means clustering algorithms. Using the extremities of the detected folds and the angle information, the taeniae coli are extracted. We also utilize the centerline of the colon for the extraction of the four major flexures on the colon surface. The projection of the centerline, evaluation of slopes, and Dijkstra’s algorithm are used to detect these flexures. These anatomical landmarks are automatically extracted using the volumetric CT images, which are then used to flatten the colon surface. These flattened colon segments can be used for future processing to detect additional landmarks and features in order to improve the supine-prone colon registration. We show the anatomical landmarks on several colon datasets. Keywordshaustral folds; taeniae coli; flexures; virtual colonoscopy