Computer-Aided Reading of Mammograms Crashes

April 4, 2007

SACRAMENTO, Calif. -- Radiologists can claim a victory over computers at reading mammograms, reported a multicenter team of investigators. Computer-aided detection resulted in 20% more biopsies, many of them false-positives.

SACRAMENTO, April 4 -- Computer-aided detection degrades the performance of screening mammography, reported a multicenter team.

In a study of more than 429,000 mammogram results, there was a significant drop in diagnostic specificity after computer-aided detection was introduced to various centers, reported Joshua J. Fenton, M.D., M.P.H., of the University of California at Davis, and colleagues, in the April 5 issue of the New England Journal of Medicine.

There was a significant increase in false-positive results and unnecessary biopsies, they added.

"In our observational study of large numbers of community-based mammography facilities and patients, the use of computer-aided detection was associated with increases in potential harms of screening mammography, including higher recall and biopsy rates, and was of uncertain clinical benefit," they wrote.

Dr. Fenton and colleagues evaluated the association between the use of computer-aided detection and the performance of screening mammography from 1998 through 2002 at 43 facilities in three states. They evaluated data on a total of 429,345 mammograms in 222,135 women, of whom 2,351 had received a diagnosis of breast cancer within a year of screening.

The authors calculated the specificity, sensitivity, and positive predictive value of screening mammography with and without computer-aided detection. They also calculated the rates of biopsy and breast-cancer detection, and at the overall accuracy of screening, which they determined by the area under the receiver-operating-characteristic curve.

In all, seven of the 43 facilities started using computer-aided detection during the study period, and the remainder of facilities served as controls.

Among the computer-using centers, the diagnostic specificity of screening mammography declined significantly from 90.2% before computers, to 87.2% after (P

"Mammography is an inherently poor, two-dimensional projectional method being used to diagnose small, three-dimensional cancers," he wrote. "It is least effective in the screening of dense breasts, which, as emphasized in another recent study, are a substantial risk factor for breast cancer."

In contrast, MRI does not expose women to radiation, and it has a high sensitivity, although lower specificity than mammography for breast cancer, he noted.

"The major problems with MRI of the breast and related magnetic resonance spectroscopy are cost and interpretive expertise," Dr. Hall wrote. "These same problems were involved with the acceptance of mammography as a screening method three decades ago. Here we go again."