CT colonography using 360-degree virtual dissection: a feasibility study.

Abstract

Using a 3D rendering technique called "virtual dissection," we sought to evaluate polyp and fold distortion using a colon phantom, estimate the polyp detection performance in humans, and estimate the added benefit of double interpretation and computer-aided diagnosis. A colon phantom containing 144 polyps of varying sizes (5-12 mm) and shapes (flat, sessile, pedunculated) was scanned. Polyp shape and distortion at virtual dissection were categorized as flame, club, pea, or bizarre. Haustral fold distortion was graded. The CT colonography examinations in 20 consecutive patients (colonoscopically proven normal findings, n = 5; polyps > or = 1 cm, n = 17 in 15 patients) were blindly reviewed by three radiologists using the virtual dissection technique. The added benefits of double interpretation and computer-aided diagnosis were tabulated. Sessile polyps appeared flame (35/48 [73%]) or pea (11/48 [23%]) in shape. Flat polyps appeared flame-shaped (31/47 [66%]) or pea-shaped (16/47 [34%]). Pedunculated polyps were flame (15/45 [33%]), club (20/45 [44%]), or pea (6/45 [13%]) in shape. Axial distortion occurred along the longitudinal axis. The sensitivities of the three observers for polyps of 1 cm or more were 16/17 (94%), 14/17 (82%), and 15/17 (88%). The specificities were 5/5 (100%), 5/5 (100%), and 4/5 (80%). Sensitivities after double interpretation and computer-aided diagnosis improved but did not reach statistical significance. Although distortion of colonic structures exists at virtual dissection, it does so in recognizable patterns, so that sensitivity for polyp detection is not compromised.