Humans Beat Computers at Mammogram Interpretation
Software designed to facilitate mammography interpreation is less accurate compared to reading by humans alone. The FDA approved computer-assisted detection in 1998 but resulting effects on assessment accuracy has yet to be examined.
A study published in the New England Journal of Medicine and supported, in part, by the National Cancer Institute compared computer-aided versus without computer support of over 425,000 mammograms at 43 different institutions during a 15 year period.
Endpoints included sensitivity, specificity and positive predictive value for each of the two methods, comparing findings for the two. Specificity decreased significantly; sensitivity and PPV both increased, but not enough to achieve statistical significance. The study also noted an increased rate of diagnostic procedures with computer detection for suspicious findings.
Related stories:
- Integrating Digital Mammography
- Mammography Screening Guidelines Challenged
- MRI for High Risk Breast Cancer
April 5, 2007 Related topics: Imaging, Quality, Safety, Errors, Diagnostic, Radiology