Computer-aided detection in digital mammography: false-positive marks and their reproducibility in negative mammograms

Abstract

There are relatively few studies reporting the frequency of false-positive computer-aided detection (CAD) marks and their reproducibility in normal cases. To evaluate retrospectively the false-positive mark rate of a CAD system and the reproducibility of false-positive marks in two sets of negative digital mammograms. Two sets of negative digital mammograms were obtained in 360 women (mean age 57 years, range 30-76 years) with an approximate interval of 1 year (mean time 343.7 days), and a CAD system was applied. False-positive CAD marks and the reproducibility were determined. Of the 360 patients, 252 (70.0%) and 240 (66.7%) patients had 1-7 CAD marks on the initial and second mammograms, respectively. The false-positive CAD mark rate was 1.5 (1.1 for masses and 0.4 for calcifications) and 1.4 (1.0 for masses and 0.4 for calcifications) per examination in the initial and second mammograms, respectively. The reproducibility of the false-positive CAD marks was 12.0% for both mass (81/680) and microcalcification (33/278) marks. False-positive CAD marks were seen in approximately 70% of normal cases. However, the reproducibility was very low. Radiologists must be familiar with the findings of false-positive CAD marks, since they are very common and can increase the recall rate in screening.