An evaluation of the impact of computer-aided detection (CAD) systems on interpreting digital mammograms found that radiologists rarely changed their diagnostic decisions after adding CAD, according to results published in the October issue of the American Journal of Roentgenology.

“Although there was room for improvement in our radiologists’ interpretations without CAD, few readers showed improved sensitivity after the application of CAD to digital mammograms,” wrote Elodia B. Cole, MS, of the Medical University of South Carolina, Charleston, and colleagues.

One of the major concerns with CAD systems is the number of false-positive marks they produce. “Each [region of interest] marked by CAD has to be looked at a second time by the radiologist, so even 0.5 false-positive marks per image (i.e., two false-positive marks per four-view mammogram) leads to an increase in interpretation time,” explained Cole and colleagues.

Despite these concerns, between 2004 and 2008 the use of CAD in screening mammography jumped from 39 percent to 74 percent. Given the widespread use of such systems and the lingering questions, the authors wanted to assess radiologist performance through a retrospective reader study in which CAD was applied to digital mammograms acquired during The American College of Radiology Imaging Network's Digital Mammographic Imaging Screening Trial (DMIST).

DMIST accrued women presenting for screening mammography from 2001 to 2003, and only cases with proven cancer status by biopsy or one-year follow-up were included in the current analysis. Two commercially available CAD systems were used: 14 radiologists interpreted cases using iCAD SecondLook and a separate group of 15 radiologists used the R2 ImageChecker Cenova system. Each group interpreted sets of 300 cases (half with cancer and half benign or normal), both with and without the assistance of CAD.

Results showed the average area under the curve was 0.71 without and 0.72 with the use of either CAD system. Sensitivity and specificity differences in performance were not statistically significant overall.

When looking at the readers individually, the CAD systems did have an impact for some. Four readers (28.6 percent) demonstrated sensitivity improvements in the iCAD group, while four readers (26.7 percent) had significant sensitivity improvements in the R2 system group.

There was an average of 0.78 marks per image for both CAD systems in the current study. Cole and colleagues noted that the high number of CAD marks per image is considered the primary reason for limited sensitivity improvements with CAD in clinical practice. They acknowledged that the 50 percent cancer prevalence rate in the current study is over one hundred times the rate experienced in normal practice. Readers were told the study was enriched with cancers, though they did not know specifically how many.

“The dismissal of so many true-positive CAD marks could be attributed in part to radiologists’ behavioral differences in clinical environments, where they see far fewer abnormal mammograms, versus laboratory environments, where from necessity they see many more cancers and benign (but still abnormal) findings,” wrote the authors.