Artificial intelligence image recognition of melanoma and basal cell carcinoma in racially diverse populations

• DOI: 10.1080/09546634.2021.1944970
• Publication Date: 2021-07-01T06:13:32Z
• Authors: Pushkar Aggarwal; Francis A. Papay
• Keywords: Basal (medicine); Tone (literature)

Artificial intelligence (AI) image recognition models have been relatively successful in diagnosing cutaneous manifestations individuals with light skin tone. However, when these are tested on the same darker or brown tone, performance of model drops due to a paucity such images available for training.The objective this study was improve AI recognizing diseases tone.Unsupervised computer darkening color preservation dermatological disease/lesion characteristics light-skinned basal cell carcinoma (BCC), and melanoma performed.Training an artificially "darkened" as compared training original "light-skinned" resulted higher sensitivity, specificity, positive predictive value, negative F1 score area under receiver-operating characteristic curve differentiating between BCC tone.Use unsupervised translation medical has potential significantly their accuracy racially diverse