Lecture notes in computer science vol:3217 pages:804-812
7th international conference on medical image computing and computer-assisted intervention - MICCAI 2004 location:Saint-Malo, France date:26-29 September 2004
A technique for Computer Aided Detection (CAD) of colonic polyps, in Computed Tomographic (CT) Colonography is presented. Following the segmentation of the colonic wall, normal wall is identified using a fast geometric scheme able to approximate local curvature. The remaining structures are modeled using spin images and then compared to a set of existing polypoid models. Locations with the highest probability of being colonic polyps are labeled as final candidates. Models are computed by an unsupervised learning technique, using a leave one out technique on a study group of 50 datasets. True positive and false positive findings were determined, employing fiber optic colonoscopy as standard of reference. The detection rate for polyps larger than 6 mm was above 85%, with an average false positive detection rate of 2.75 per case. The overall computation time for the method is approximately 6 minutes. Initial results show that Computer Aided Detection is feasible and that our method holds potential for screening purposes.
Kiss G., Van Cleynenbreugel J., Marchal G., Suetens P., ''Computer aided detection in CT colonography, via spin images'', Lecture notes in computer science, vol. 3217, pp. 804-812, 2004 (Proceedings 7th international conference on medical image computing and computer-assisted intervention - MICCAI 2004, part II, September 26-29, 2004, Saint-Malo, France).