Researchers tackle false-positives in CAD

Mar 16, 2004

Computer-aided detection (CAD) is a promising technology that is changing the way many breast imaging facilities interpret mammograms. But CAD still needs some fine-tuning, in particular to address a high false-positive rate that leaves mammographers relying heavily on biopsy to determine whether CAD-flagged areas really are malignant.

Dear AuntMinnie Member,

Several articles we’re featuring this week in our Women’s Imaging Digital Community deal with CAD in general, and more specifically with the false-positive dilemma. Our first article, by staff writer Shalmali Pal, tracks the efforts by a joint team from the U.S and Israel to analyze the way CAD reviews microcalcifications to see whether a new data-analysis algorithm could provide high sensitivity with a lower biopsy rate.

Another study, in the same article, tracks the use of a CAD-based lesion-analysis program to improve the characterization of suspicious areas in the BI-RADS 3 category, a tricky gray area that includes lesions that are probably benign but that require follow-up anyway. The CAD application was able to correctly upgrade suspicious lesions to a more serious BI-RADS category with 90% sensitivity, according to the authors.

Finally, in a report from the just-concluded European Congress of Radiology, staff writer Eric Barnes describes the experience German researchers had in test-driving a new version of one of the leading CAD software programs. While sensitivity was only slightly improved, they found that the program was effective in reducing the false-positive rate significantly.

Get all the details in our Women’s Imaging Digital Community, at womens.auntminnie.com.