AJR:192, June 2009
Stuart A. Taylor, Johrr Brittenden, James Lenton, Hannah Lambie, Anthony Goldstone, Peter N.Wylie, Damian Tolan, David Burling, Leslie Honeyfield, Paul Bassett, Steve Halligan
OBJECTIVE. The objective of our study was to investigate whether an increasing num¬ber of computer-aided detection (CAD) false-positives decreases reader sensitivity, specific¬ity, and confidence for nonexpert readers of CT colonography (CTC).
MATERIALS AND METHODS. Fifty CTC data sets (29 men; mean age, 65 years), 25 of which contained 35 polyps > 5 mm, were selected in which CAD had 100% polyp sen¬sitivity at two sphericity settings (0 and 75) but differed in the number of false-positives. The data sets were read by five readers twice: once at each sphericity setting. Sensitivity, speci¬ficity, report time, and confidence before and after second-read CAD were compared using the paired exact and Student's / test, respectively. Receiver operating characteristic (ROC) curves were generated using reader confidence (1-100) in correct case classification (normal or abnormal).
RESULTS. CAD generated a mean of 42 (range, 3-118) and 15 (range, 1-36) false-posi¬tives at a sphericity of 0 and 75, respectively. CAD at both settings increased per-patient sen¬sitivity from 82% to 87% (p = 0.03) and per-polyp sensitivity by 8% and 10% for a sphericity of 0 and 75, respectively (p < 0.001). Specificity decreased from 84% to 79% (sphericity 0 and 75, p = 0.03 and 0.07). There was no difference in sensitivity, specificity, or reader con¬fidence between sphericity settings (p = 1.0, 1.0, 0.11, respectively). The area under the ROC curve was 0.78 (95% CI, 0.70-0.86) and 0.77 (0.68-0.85) for a sphericity of 0 and 75, respec¬tively. CAD added a median of 4.4 minutes (interquartile range [IQR], 2.7-6.5 minutes) and 2.2 minutes (IQR, 1.2-4.0 minutes) for a sphericity of 0 and 75, respectively (p < 0.001).
CONCLUSION. CAD has the potential to increase the sensitivity of readers inexperi¬enced with CTC, although specificity may be reduced. An increased number of CAD-gener-ated false-positives does not negate any beneficial effect but does reduce efficiency.