[CT colonography: evaluation of two 3D algorithms in a screening population]

The purpose of this multicenter study was to compare a dissection display and an endoluminal display for CT colonography (CTC) by means of detection rates and evaluation time in a screening collective.

4 blinded readers evaluated CTC datasets from 42 patients with 55 endoscopically confirmed polyps. The datasets were read in a randomized order using two different 3D visualization methods (endoluminal view vs. dissection display; EBW 2.0.1, Philips Medical Systems, Best/NL). Patients underwent cathartic cleansing as well as stool and fluid tagging. All readers except one were experienced in performing CTC. The per-lesion/per-patient sensitivity, per-patient specificity, and evaluation time were calculated.

The overall per-lesion sensitivity using the dissection display (and endoluminal view) was 60% (53 %) for reader 1, 58% (60%) for reader 2, 67% (71%) for reader 3 and 55% (58%) for reader 4. The per-patient sensitivity using the dissection display (and endoluminal view) was 85% (85%) for reader 1, 80% (85%) for reader 2, 95% (90%) for reader 3 and 80% (80%) for reader 4. The per-patient specificity was 68% with dissection view (77% endoluminal view) for reader 1, 82% (82%) for reader 2, 59% (59%) for reader 3 and 82% (73%) for reader 4. The experienced readers were significantly faster using the perspective-filet view.

Using a dissection display of CTC datasets does not result in superior detection rates for polyps if datasets are stool and fluid-tagged. 3 out of 4 readers evaluated the datasets significantly faster with the dissection display.