Mammography computer-aided detection (CAD) Image Source: Definiens

Computer-aided detection (CAD) significantly increases the time it takes radiologists to review mammograms, while only slightly bolstering their diagnostic confidence, according to a study published in the October issue of Radiology.

“The consensus thus far is that CAD provides some improvement in cancer detection, albeit with increased time and cost,” Phillip M. Tchou, PhD, of University of Texas M.D. Anderson Cancer Center in Houston, and colleagues wrote.

The purpose of the study was to analyze how CAD affected radiologists’ reading times, diagnostic decisions and confidence in their mammography readings. Five radiologists examined 267 mammograms, with the average reviewer holding 17 years experience. 

Radiologists first read images without CAD and decided if the images were negative or if they merited callback for further testing. The radiologists’ reading time as well as his confidence (on a scale from 1 to 10) in the decision was recorded, after which the physician reread the same mammogram with CAD markings. The CAD review time and updated confidence (on a scale from 1 to 10) were then recorded.

The authors found that CAD increased the mean reading time of the five radiologists by 19 percent, from 18 to 23 seconds. The study also showed that CAD increased reading time most for the more experienced doctors. Tchou and colleagues hypothesized that the discrepancy resulted from increased caution with experience or younger physicians feeling more comfortable with the computerized system. The authors also reported that masses that were marked by CAD increased physicians’ reviewing times more than CAD-marked calcifications.

Thirty-five patients were recalled for further testing; four of the 35 were called back only after the physician used CAD to re-examine the images. One patient was initially due for further testing but upon rereading with CAD, the radiologist decided a callback was unnecessary. In total, the use of CAD changed the radiologists’ decisions in two percent of readings. CAD increased the recall rate 11 percent among the five readers.

Out of the four patients in the study who were diagnosed with cancer, one would not have been called back without CAD.
 
In 14 percent of the readings, reviewers reported that they felt more confident about their findings after re-examining images using CAD, while in eight percent of cases physicians responded that CAD made them less confident in their original decisions.

The researchers acknowledged several limitations to their findings. “First, the study length and the number of cases were small compared with other CAD studies. As a result, our study had greater statistical error,” Tchou and colleagues noted. The researchers also called their experimental setting partly “artificial,” since radiologists would not normally read a study twice and publicly announce their decision and confidence. 

Still, the study reinforced previous findings—CAD provides some advantage in detecting cancer, but adds time and money to the process. The authors expressed their future aims to explain how CAD affects physicians’ confidence and to minimize additional CAD reading time while preserving the technology’s effectiveness.