Objective: To determine the feasibility of a computer-aided detection (CAD) algorithm as the “first reader” in computed tomography colonography (CTC).
Methods: In phase 1 of a 2-part blind trial, we measured the performance of 3 radiologists reading 41 CTC studies without CAD. In phase 2, readers interpreted the same cases using a CAD list of 30 potential polyps.
Results: Unassisted readers detected, on average, 63% of polyps ≥10 mm in diameter. Using CAD, the sensitivity was 74% (not statistically different). Per-patient analysis showed a trend toward increased sensitivity for polyps ≥10 mm in diameter, from 73% to 90% with CAD (not significant) without decreasing specificity. Computer-aided detection significantly decreased interobserver variability (P = 0.017). Average time to detection of the first polyp decreased significantly with CAD, whereas total reading case reading time was unchanged.
Conclusion: Computer-aided detection as a first reader in CTC was associated with similar per-polyp and per-patient detection sensitivity to unassisted reading. Computer-aided detection decreased interobserver variability and reduced the time required to detect the first polyp.
From the *Department of Radiology, Stanford University Medical Center, Stanford, CA; †Stanford Medical School, Stanford, CA; ‡Department of Radiology, University of California at San Francisco, San Francisco, CA; §San Francisco Veterans Administration Medical Center, San Francisco, CA; ║Veterans Affairs Palo Alto Health Care System, Palo Alto, CA; ¶Department of Radiology, University of Vienna, Vienna, Austria; and #Department of Medicine, Gastroenterology, Stanford University Medical Center, Stanford, CA.
Funded in part by the National Institutes of Health (1R01 CA72023, 1P41 RR09784-01) and Society for Computed Body Tomography and Magnetic Resonance. Aravind Mani was supported by the Stanford Medical School Scholars Program. Additional support received from the Packard Foundation (Los Altos, CA), Lucas Foundation (Menlo Park, CA), and Phil N. Allen Trust (Menlo Park, CA).
Presented at the 88th Scientific Assembly and Annual Meeting, Radiologic Society of North America, Chicago, IL, December 1–6, 2002 (Radiology. 2002;225(P):304).
Reprints: Dr Christopher F. Beaulieu, Department of Radiology, Stanford University Medical Center, MC 5105, 300 Pasteur Drive, Stanford, CA 94305 (e-mail: firstname.lastname@example.org).