Colonic polyp characterization and detection based on both morphological and texture features

In this paper, a method is presented for detection of colonic polyps using both morphological and texture features of the colon wall from the abdominal computed tomography (CT) images. This method consists of two steps. In the first step, suspicious patches of the colon wall are quickly identified by utilizing special local and global geometrical information of the segmented inner-wall mucosa layer. In the second step, a special mapping strategy is employed to identify the growing region of each suspected polyp. Then its internal CT density textures are extracted and quantitatively analyzed based on an assumed ellipsoid polyp model. Finally, both the extracted texture and morphological information are applied to eliminate the false positives from the identified suspicious patches. With all the extracted geometrical, morphological and texture features, this presented computer-aided detection (CAD) method has demonstrated a significant improvement in detection of the colonic polyps over the previously-reported geometry-based CAD methods for CT-based virtual colonoscopy.