Clinical evaluation of assistant diagnostic system for mammograms using the auto-analyzing method

Clinical usefulness of a newly developed auto-analyzing system of mammograms to search the abnormal dense areas such as cancer lesions was studied. 345 breasts including 86 cancers, 52 mastopathies, 18 fibroadenomas, 13 cysts, 4 lipomas and 172 normals were analyzed and compared with final diagnosis. 84 of 86 cancer lesions were correctly pointed out (sensitivity: 97.9%). Based on the gradient of the density, the ranking of the cancer or calcified lesions among suspected regions in a mammogram was decided. 55 lesions of them (64.0%) were ordered at the first rank and 83 (96.5%) were in the upper three ranks. 68 of 74 breasts with calcifications were correctly diagnosed and 48 of them were ranked in the upper three ranks. Although further improvement of the system is needed, this system will be a useful assistance for screening the breast cancer in the analyses of mammograms.