Algorithm makes polyp surveillance nearly automatic on CTC

An algorithm for automatic endoluminal coregistration at initial CT colonography can successfully help predict polyp location at subsequent CT colonography, according to a study published online July 4 in Radiology.

Typically, when CT colonography is used for polyp surveillance, it’s a matter of a radiologist comparing an initial CT colonography study in which one or more polyps was detected to a subsequent study, manually matching the polyps for measurement.

“However, matching polyps from serial CT colonographic examinations can be challenging and time consuming, especially because such polyps are usually small,” wrote study authors Emma Helbren, MBChB, FRCR, of the Centre for Medical Imaging at University College London, and colleagues.

Computer algorithms that coregister endoluminal surface location on CT colonographic images have been proposed as a way to more easily match polyps, but little research had previously been done to evaluate their effectiveness.

Helbren and colleagues used a registration algorithm designed to coregister the coordinates of endoluminal colonic surfaces on images from prone and supine CT colonographic acquisitions to match polyps in sequential studies in patients undergoing polyp surveillance. A total of 26 patients with 35 polyps were selected for the study. The algorithm was tested by a longitudinal method, in which polyp coordinates from initial prone and supine acquisitions were used to plot expected polyp location on a follow-up study, and the consistency method, in which polyp coordinates from an initial supine acquisition alone were used to identify polyps in follow up prone and supine acquisitions.

Results showed coregistration was achieved for all 35 polyps using both testing methods. Mean standard deviation of registration error for the longitudinal method was 17.4 mm, and for the consistency method was 26.9 mm. There was no significant difference in registration error when prone and supine acquisitions in the same study were compared, according to the authors.

By bringing the observer immediately to within a couple centimeters of where a polyp would be if it is still present on a second study, Helbren and colleagues noted that such algorithms can aid in the efficient identification of known polyps in patients undergoing CT colonography surveillance. “It may be especially helpful for locating small polyps and also for the accurate identification and registration of individual polyp locations when multiple polyps are present, a situation when conventional unassisted coregistration can be particularly challenging and time consuming for the radiologist.”