Asking for help: Interactive CAD bests unprompted CAD assistance

Radiologists’ interpretations of mammograms could be improved if an interactive computer-aided detection (CAD) system is used, as opposed to traditional CAD systems with prompts designed to prevent oversights, according to a study published online ahead of print in Radiology.

“Breast cancers might be detected earlier in screening without an increase in false-positive recalls if CAD results for masses and architectural distortions can be queried interactively and are accompanied by a suspiciousness score,” wrote Rianne Hupse, MSc, of Radboud University Nijmegen Medical Centre, Nijmegen, the Netherlands, and colleagues.

Typically, CAD systems are used to prevent perceptual oversight of abnormalities by automatically prompting users. However, the authors explained that while CAD’s ability to detect microcalcification is high, there is less agreement on the usefulness of CAD for detection of masses and architectural distortions. Previous research has determined that reader performance could be improved by combining user scores with CAD marks.

“These results motivated us to develop a CAD system that helped radiologists interpret suspicious regions, rather than to assist with the initial detection of these regions,” wrote the authors. “In our interactive system, CAD marks are only displayed ‘on demand’ for queried regions with a suspiciousness score.”

To test the effectiveness of this interactive CAD system, Hupse and colleagues conducted a retrospective observer study featuring nine certified screening radiologists and three residents who were trained in breast imaging. They read a total of 200 studies—63 containing at least one screen-detected mass, 17 false-negative studies, 20 false-positive studies and 100 normal studies. Studies were read once with CAD prompts, and once with the interactive CAD system.

Results showed the average partial area under the location receiver operating characteristic curve with unaided reading was 0.57. While this was unaffected by prompts, it increased to 0.62 with interactive CAD, a statistically significant improvement.

“The reading process of the proposed interactive system was not disrupted by appearance of false-positive prompts at unexpected locations, as was the case in the conventional system, which may have led to these results,” wrote the authors. “In the interactive mode, marks remained hidden unless regions that corresponded were probed. Because most radiologists probed only a limited number of regions, and only those that they were interested in, fewer false-positives were displayed.”

Hupse and colleagues did note a large variance in the effect of interactive CAD among the 12 readers, based largely on experience. Those with more experience had higher unaided performance, thus reducing or eliminating the benefit of CAD.