Automatic segmentation, tissue characterization, and rapid diagnosis enhancements to the computed tomographic colonography analysis workstation

Abstract

An image processing system developed to support examination of computed tomographic colonoscopy (CTC) was developed in 1995. The clinical viability of CTC is enhanced by the solution of several technical problems. These problems include the limited detectability of sessible polyps and difficulties in discrimination between polypoid masses and retained stool. CTC is also made more feasible by simplifying the required colon preparation and reducing the time required to analyze scan results. Each of these challenges have been addressed by enhancements to the CTC analysis workstation software. Endoluminal volume rendering has been enhanced by the addition of automatic segmentation to facilitate analysis of colon segments, which contain tagged liquid stool. By automating this function, the system is able to process scans that are acquired following a wide variety of colon preparation protocols. Similar approaches have been used to identify retained stool. Automatic tissue characterization has also been incorporated into the volume rendering routines to help identify and diagnose polypoid masses. These enhancements have improved the quality of CTC interpretation, while reducing the time required to perform the analysis. This time reduction was necessary to reduce the cost of CTC enough to make it viable for asymptotic population screening. To date, over 150 patient examinations have been performed using this new technique. A recent blinded, prospective study reporting the results from two independent observers has been presented. The technique is feasible, reliable, and has been implemented clinically with results reported within 1 hour of the examination.