CAD has little effect on radiologists' diagnoses

Radiologists rarely change their diagnoses after using computer-aided detection (CAD) systems with digital mammography, according to a study published this month in the American Journal of Roentgenology.

Ever since CAD for mammography was introduced, there have been concerns about the technology's benefits--especially considering the high number of false positives it produces. For example, research published last year in the Annals of Internal Medicine found that while the use of CAD during screening mammography among Medicare patients was associated with the earlier detection of cancer, it also increased the risk of false positives.

Despite such concerns, according to Elodia Cole, of the Medical College of South Carolina, and colleagues, the use of CAD in screening mammography increased from 39 percent to 74 percent between 2004 and 2008.

The question, then, was how much the increased use impacted radiologist performance. The researchers analyzed the impact of CAD on radiologists performance on digital mammograms acquired during the American College of Radiology Imaging Network's Digital Mammographic Imaging Screening Trial. Although CAD wasn't used during that trial, the researchers believed it offered a large database in which CAD could be applied retroactively.

In the study, 14 radiologists interpreted a set of 300 cases (half cancer, half benign) with and without the use of CAD. The researchers found there were no significant differences in specificity and sensitivity performance with or without the use of CAD. In addition, the use of CAD didn't influence the radiologists to change their cancer ratings.

According to research published in AJR in 2012, CAD improved the likelihood for radiologists to identify cancer that initially went undetected during a screening. The study's authors said that radiologists failed to recognize a correct computer prompt in 70 percent of missed cancer cases.

To learn more:
- see the study in the American Journal of Roentgenology