Are unnecessary follow-up procedures induced by computer-aided diagnosis (CAD) in mammography? Comparison of mammographic diagnosis with and without use of CAD

Objective: To evaluate the rate of unnecessary follow-up procedures recommended by radiologists using a CAD-system. Materials and methods: 185 patients (740 images) were consecutively selected from three groups (36 histologically proven cancers=group 1; 49 histologically proven benign lesions=group 2 and 100 screening cases (4 years-follow up=group 3). Mammograms were evaluated by a CAD system (Second Look[reg], CADx, Canada). Five blinded radiologists assessed the images without/with CAD outputs. Diagnostic decisions were ranked from surely benign to surely malignant according to BIRADS classification, follow-up procedures were recommended for each observed lesion (a, screening; b, short interval follow-up examination in 6 months; c, pathologic clarification). Results: CAD-system detected 32/36 cancers (88.9%) (FP-rate: 1.04 massmarks and 0.27 calcmarks/image). The following values were reached by all observers without/with CAD in the mean: Sensitivity 80.6/80.0%, specificity 83.2/86.4%, PPV 53.1/58.1%, and NPV 94.6/94.7%. Observers described a similar number of additional lesions without/with the use of CAD (325/326). Whereas the number of unnecessary short-time follow up recommendations increased in all case-subgroups with CAD: 40.8/42.9% (group 1), 35.6/38.1% (group 2), 44.7/46.8% (group 3), respectively, the number of recommended biopsies decreased in all subgroups: group 1: 34.7/27.1%; group 2: 47.4/41.5%, group 3: 33.3/22.0%, respectively. Conclusion: In this rather small population additional usage of CAD led to a lower rate of unnecessary biopsies. The observed decrease of recommended unnecessary biopsies due to the usage of CAD in the screening group suggests a potential financial benefit by using CAD as diagnostic aid