Aim: to compare the diagnostic efficacy of generation I and II computer aided detection (CAD) systems for mammography of our own design using the large set of unselect ed mammography images obtained in a routine clinical practice settings. Material and methods. Both CADs were tested on the set of 1532 mammography images of 356 women with confirmed breast cancer (BC). We assessed their value in the detection of suspicious areas with various characteristics located on the different density background. Size of BC lesions varied from 4 to 35 mm (mean – 13,4 ± 6,3 mm). We excluded BC representing only with microcalcification clusters from this analysis, because this task is solved using the separate universal module compatible with both CADs.Results. For I and II generation CADs we obtained the following results: detection of small nodular BCs (≤10 mm) – 41 of 52 (78.85%) and 48 of 52 (92.31%; p > 0.05), respectively; detection of BCs visible as asymmetric areas – 18 of 18 (100%) and 13 of 18 (72.2%; p > 0.05), respectively; detection of only partially visible masses – 15 of 18 (83.3%) and 17 of 18 (94.4%; p > 0.05); detection of lesions poorly visible or invisible on standard mammography images due to the high density background (C-D types according to the ACR 2013 classification) – 9 of 16 (56.3%) and 7 of 16 (70.0%; p = 0.046). Total detection rate was 88.76% (316 of 356 cases) – for CAD I and 90.73% (323 of 356 cases; р > 0.05) – for CAD II. Mean false positive marks rate was 1.8 and 1.3 per image, respectively, – for ACR А-В images and 2.6 and 1.8 per image, respectively – for ACR C-D images (p < 0.05).Conclusion. Generally the diagnostic value of CAD II is not inferior that of CAD I in all analyzed situations, except the poorly visible or invisible lesions on the dense breast background. Moreover, CAD II is probably superior CAD I in the detection of spiculated small masses. The rate of false positive marks was significantly higher for CAD I.