Dermatologist-level classification of skin cancer with deep neural networks

“…These results differ from most theoretical and retrospective studies in which AI accuracy is usually equal to or higher than that of clinicians [23,26,27,38], and are consistent with the few existing prospective and real-world studies [50]. In addition, it is of relevance that the speci city of the application of AI in dermatologic imaging was very close to 1, which suggests that it is a useful tool for application in routine clinical practice as a CDST Moreover, the fact that the diagnostic accuracy metrics increase with the Top-3 and Top-5 assessment is consistent with the usefulness in differential diagnosis, a fact already pointed out by Muñoz-López C et al in their study [50].…”

“…Our results showed that the accuracy of examining FSs by pathology experts was generally higher than that of the 18 models. This result is substantially different from the results in most articles 12 , 23 , 24 . The possible reasons for the high degree of homogeneity in the FS diagnosis of sentinel lymph nodes by our pathologists may be as follows: first, the diagnosis of lymph node metastasis is a basic and essential skill for cancer pathologists, as they have been heavily trained on this skill; second, the 5 senior pathologists in this group are all breast cancer specialists and have extensive experience; third, the differences in years of practice may have little effect on the consistency of lymph node diagnosis in the results of our review of 160 randomly chosen FSs.…”

“…These results are consistent with prior findings of high diagnostic discrimination in AI dermatology systems [ 1 , 2 , 3 , 6 ].…”