Computer-assisted evaluation of mammography images. Initial clinical experiences

Preceding studies have shown that a second independent reviewer of conventional mammographies increases the detection rate of features typical for malignancy by up to 15%. METHODS: In order to test a computer-aided diagnostic (CAD) system (ImageChecker, R2 Technology, USA) for the detection of pathologic criteria in conventional mammography, 96 mammographies were retrospectively evaluated using ImageChecker. Thirty-five of these mammographies had been diagnosed as not showing pathologies, and 61 had depicted histologically confirmed malignancy. RESULTS: Detecting 41 of 61 breast malignancies, ImageChecker showed a diagnostic sensitivity of 70.5%. All malignancies accompanied by microcalcifications were identified by ImageChecker, whereas 18 cases characterized by parenchymal opacity without microcalcifications were not marked. On the average, 1.95 markers per image were set, giving a total of 187 markers in this study. 63% of all markers showed normal tissue and were thus false positive. CONCLUSIONS: Pathologic parenchymal opacities in mammography are a well-known problem for all CAD systems in use. Despite this major drawback, even now ImageChecker can provide tremendous support in routine interpretation of conventional mammographies.