Abstract
We have been developing a computer-aided diagnosis (CAD) system for mammograms and have reported our evaluation studies. The purpose of this study is to evaluate the potential effectiveness of the system in detecting masses and clustered microcalcifications by comparing 142 doctors' and the CAD system's detection results with 200 mammograms for 100 patients employed in doctors' short training course. The sensitivity in the detection of mass was 85% with 2.1 false positives (FPs) per image and that of clusterd microcalcifications was 100% with 0.8 FPs per image for this database. We analyzed the data based on the interpretation experiences of the doctors. We found that our CAD system was able to detect some of the candidates which were missed by some doctors. Especially, the system shows the possibility of the stable assistance in detection for the physicians with less experience of the reading. In conclusion, these results imply that our CAD system has a potential to improve the accuracy of the image interpretation.
