Are CAD Marks in Full-Field Digital Mammography Consistently Reproducible?

The objective of a recent study was to evaluate the sensitivity and reproducibility of CAD on digital mammography and follow-up digital mammography.

The study has concluded that computer-aided detection (CAD) in full-field digital mammography offers high sensitivity and reproducibility.

Methodology
The retrospective review included 95 cancers, with a total of 190 mammograms with cancer and 174 mammograms without evidence of cancer. The mean time interval between two consecutive mammographic examinations was 23 days.

There were 70 malignant masses and 49 clusters of malignant calcifications. Review of sensitivity, reproducibility, and false-positive rate of CAD was performed.

Results of the Study
The sensitivity of the CAD for masses on initial digital mammography was 91%, and that of follow-up mammography was 89%. These rates were not statistically significant. The sensitivity of CAD for masses in fatty breasts was 100%, and was statistically higher than in dense breasts.

The sensitivity of CAD for microcalcifications on both initial and follow-up digital mammography was 100% and was not affected by density of the breast.

The rate of false-positive marks was 0.29 marks per image for initial mammography, and the rate of false-positive marks was 0.27 marks per image for follow-up mammography.

No statistical difference was found in the false-positive marks between masses and microcalcifications. The false-positive mark rate was not affected by the density of the breast and was not statistically different between masses and microcalcifications.

Overall mark-based reproducibility was 80% for true-positive mass marks, and 92% for microcalcifications. The overall mark-based reproducibility for false-positive marks was 9%.

Conclusions
The sensitivity of CAD for malignant masses and microcalcifications remain high on serial digital mammography with a high reproducibility rate.

Reviewer’s Comments
Many radiologists have incorporated CAD into their daily practice to reduce the false-negative rate of screening mammography. CAD has been shown to increase the sensitivity of mammography and to improve the cancer detection rate. Although a recent retrospective multi-institutional review suggested a decrease in specificity and positive predictive value without a significant improvement in the cancer detection rate, the authors showed reproducible high sensitivity of CAD in full-field digital mammography.

Author: Bokran Won, MD

Reference
Kim SJ, Moon WK, et al. Computer-Aided Detection in Full-Field Digital Mammography: Sensitivity and Reproducibility in Serial Examinations. Radiology; 2008; 206 (January): 71-80: