Computer-assisted detection may give radiologists an edge in diagnosing cancer in dense breasts. Breast density does not affect the detection rates of CAD achieved with fatty breasts, according to a study presented by Dr. Rachel F. Brem at the RSNA meeting Thursday.

CAD was adept at finding microcalcifications in dense breasts, but it did less well at identifying subtle lesions embedded in dense tissue, said Brem, director of breast imaging and intervention at George Washington University.

"Our data seem to suggest that CAD may be particularly helpful in patients with dense breasts and that the sensitivity for the detection of breast cancer using mammography with CAD on BI-RADS 3- and 4-density breasts was higher than with mammography alone, physical examination, or ultrasound," Brem said.

Researchers used a CADx SecondLook system equipped with version 4 software to evaluate 906 normal and 153 positive screening mammograms from 18 facilities. All mammograms were digitized for the study. Experienced radiologists interpreted the films with and without the aid of CAD evaluation. BI-RADS 1 and 2 cases were classified as fatty breasts, and BI-RADS 3 and 4 cases were classified as dense breasts. The sensitivity of unassisted mammography, bilateral ultrasound, and physical examination was also calculated.

The cancers themselves were categorized as masses or microcalcifications, Brem said. Cancers that had both masses and microcalcifications were classed according to the primary characteristics of the tumor. The sensitivity and false-positive rate of the CAD system were compared for dense and nondense breasts.

No significant difference was seen in the readers' ability to detect cancer in dense and fatty breasts when CAD was used. The readers using CAD achieved 90% sensitivity for fatty breasts and 88% sensitivity for dense breasts, according to Brem.

The differences in the sensitivity rates for CAD in dense and nondense breasts were statistically significant (p<0.02): 89% and 83%, respectively. Statistical significance (p<0.15) was also found in the difference between the false-positive rates in the two tissue classes. An average of 0.66 false positives per film were seen in the nondense breasts compared with 0.83 false positives in dense breasts.

Using CAD, the readers did better at identifying microcalcifications than masses in dense breasts, achieving average sensitivity of 95% in dense breasts and 93% in nondense breasts.

In BI-RADS 1 fatty breasts, mammography combined with CAD was less sensitive than mammography alone, bilateral ultrasound, or physical examination. For fatty breasts, the average sensitivity of mammography alone was 98% compared with 88% for mammography and CAD.

The sensitivity rates for BI-RADS 2 were 91% for CAD mammography, 83% for mammography alone, 66% for ultrasound, and 32% for physical exam. For BI-RADS 3 dense breasts, they were 81% for CAD mammography, 64% for mammography, 81% for ultrasound, and 29% for physical examination. For extremely dense breasts, the breakdown was 83% for CAD mammography, 48% for mammography, 76% for ultrasound, and 35% for physical examination.