Improvement of radiologists' characterization of mammographic masses by using computer-aided diagnosis: An ROC study

PURPOSE: To evaluate the effects of computer-aided diagnosis (CAD) on radio logists' classification of malignant and benign masses seen on mammogram. MATERIALS AND METHODS: The authors previously developed an automated comput er program for estimation of the relative malignancy rating of masses. In t he present Study, the authors conducted observer performance experiments wi th receiver operating characteristic (ROC) methodology to evaluate the effe cts of computer estimates on radiologists' confidence ratings. Six radiolog ists assessed biopsy-proved masses with and without CAD. Two experiments, o ne with a single view and the other with two views, were conducted. The cla ssification accuracy was quantified by using the area under the ROC curve, A(z). RESULTS: For the reading of 238 images, the A(z) value for the computer cla ssifier was 0.92. The radiologists' A(z) values ranged from 0.79 to 0.92 wi thout CAD and improved to 0.87-0.96 with CAD. For the reading of a subset o f 76 paired views, the radiologists' A(z) values ranged from 0.88 to 0.95 w ithout CAD and improved to 0.93-0.97 with CAD. Improvements in the reading of the two sets of images were statistically significant (P = .022 and .007 , respectively). An improved positive predictive yalue as a function of the false-negative fraction was predicted from the improved ROC curves. CONCLUSION: CAD may be useful for assisting radiologists in classification of masses and thereby potentially help reduce unnecessary biopsies.