Progressive region-based colon extraction for computer-aided detection and quantitative imaging in cathartic and non-cathartic CT colonography

摘要

Automated colon extraction is an important first step for applications of computer-aided detection (CADe) and quantitative imaging in computed tomographic colonography (CTC). However, previously developed colon extraction algorithms have various limitations. We developed a new fully automated progressive region-based (PRB) method to extract the complete region of colon from CTC images while minimizing the presence of extra-colonic components. The method can be used to provide the target region for CADe as well as to provide quantitative imaging information about the colon. In the method, extra-colonic components are excluded from the abdominal region by use of anatomy-based extraction methods, visible lumen regions of the colon and small bowel are decomposed into material-based subregions, a colon pathway is tracked from the anus to cecum by use of algorithms of progressively increasing complexity using anatomy-based landmarks, segmental features, and region-based tracking algorithms, and the extracted lumen region is perfected into a complete lumen by use of a discrete level-set algorithm. The method was tested with 15 challenging cathartic and non-cathartic fecal-tagging CTC cases. Preliminary results indicate that the PRB method can outperform our previously developed centerline-based tracking method in colon extraction.