Computer-Aided Detection No Better Than Unaided Radiologists for Mammography Interpretation

Laurie Barclay, MD

February 03, 2004

Feb. 3, 2004 — Computer-aided detection (CAD) was not more effective than radiologists' unaided interpretation of mammography, according to the results of an analysis of outcomes databases published in the Feb. 4 issue of the Journal of the National Cancer Institute. The editorialists call this an important contribution and comment on the large expense without improved outcomes.

"Computer-aided diagnosis (discrimination) systems are currently being developed to help radiologists determine whether an identified suspicious region is likely to represent a benign or a malignant finding," write David Gur, ScD, and colleagues from the University of Pittsburgh in Pennsylvania. "The hope is that these CAD systems will improve the sensitivity of mammography without substantially increasing mammography recall rates, in addition to possibly decreasing inter-reader variability."

At Magee-Womens Hospital of the University of Pittsburgh Medical Center from 2000 to 2002, 24 radiologists interpreted 56,432 mammograms without the use of CAD and 59,139 mammograms using the CAD system.

Recall rates (approximately 11%) and breast cancer detection rates (approximately 3.5 cases detected per 1,000 screening mammograms) were not statistically different between groups. These results were very similar among a subgroup of the seven radiologists who interpreted the most studies.

Study limitations include low number of breast cancers, high inter-reader variability, inability to assess intra-reader variability, lack of adjustment for learning effect or motivation, and use of conventional film rather than digital mammography.

"Actually observed changes in our practice were substantially lower than expected," the authors write. "This is not to say that the use of CAD would not be beneficial or cost-effective in other practices. Rather, we suggest that, at its current level of performance, CAD may not improve mammography recall or breast cancer detection academic practices similar to ours that employ specialists for interpreting screening mammograms."

Public Health Service grants helped support this study.

In an accompanying editorial, Joann G. Elmore, MD, from the University of Washington in Seattle, and Patricia A. Carney, PhD, from Dartmouth Medical School, discuss other study limitations. The data did not take into account the characteristics of the women in the study, and follow-up was limited for women diagnosed with breast cancer.

They also discuss the financial and medicolegal implications of CAD. "[These results] highlight the importance of continued investigation of new screening technologies after they have received [Food and Drug Administration] approval," they write. "As more studies on CAD in the community setting are published and as CAD systems improve, we hope that women's lives can ultimately be saved by this new technology."

J Natl Cancer Inst. 2004;96:162-163, 185-190

Reviewed by Gary D. Vogin, MD


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: