Computer-Aided Detection of Non-polypoid Flat Lesions in CT Colonography: Observer Performance Study
To evaluate the effect of computer-aided detection (CADe) on the performance of human readers in the detection of non-polypoid flat lesions from a large computed tomography (CT) colonography population. A total of 153 cathartic CT colonography cases, including 45 colonoscopy-confirmed, morphologically flat lesions, were sampled from a European multi-center CT colonography trial for asymptomatic patients at increased risk of colorectal cancer. Two readers (expert and non-expert) reviewed the 153 CT colonography cases and recorded all detected lesions using primary 3D interpretation and a CADe second-read paradigm. There were 17 patients with 18 flat lesions ≥10 mm in size and 17 patients with 27 flat lesions 6 – 9 mm in size. For the flat lesions ≥10 mm, per-patient sensitivities of the expert reader for unassisted and CADe-assisted readings were 59% [95% CI: 36–78%] and 71% [47–87%], respectively, whereas those of the non-expert reader were 41% [21–65%] and 47% [37–59%], respectively. For 6-9 mm flat lesions, the corresponding per-patient sensitivities of the expert reader were 59% [36–78%] and 76% [53–89%], respectively, whereas those of the non-expert were 47% [37–59%] and 82% [59–93%]. The results indicate that the use of CADe can increase the sensitivity of human readers in the detection of flat lesions in a screening setting.
KeywordsCT colonography flat lesions non-polypoid lesions computerassisted detection observer study
Unable to display preview. Download preview PDF.
- 1.ACS. Cancer Facts & Figures 2007. American Cancer Society (2007) Google Scholar
- 5.Levin, B., Lieberman, D.A., McFarland, B., et al.: Screening and surveillance for the early detection of colorectal cancer and adenomatous Polyps, 2008: a joint guideline from the American Cancer Society, the US Multi-Society Task Force on Colorectal Cancer, and the American College of Radiology. CA. Cancer. J. Clin. 58, 130–160 (2008)CrossRefGoogle Scholar