The performance of an artificial intelligence-based computer vision mobile application for the image diagnosis of genital dermatoses: a prospective cross-sectional study.

Background: There is a huge demand-supply gap between the incidence of genital dermatoses (including sexually transmitted infections and non-venereal genital dermatoses) and physicians trained to manage them.

Objectives: To find out the performance of an artificial intelligence (AI)-based mobile application in the image diagnosis of genital dermatoses, and to compare it with primary care physicians (PCPs) and dermatologists.

Methods: Photos of the genital diseases of consecutive patients presenting to the STD and genital diseases clinic were included. The gold standard diagnosis was established by the consensus of two certified dermatologists after examination and one positive investigation. Image diagnoses by the DermaAId application, two PCPs, and two dermatologists were recorded and compared to the gold standard diagnosis and to each other.

Results: A total of 257 genital disease images, including 95 (37.0%) anogenital warts, 60 (22.2%) lichen sclerosus, 20 (7.8%) anogenital herpes, 15 (5.8%) tinea cruris, 14 (5.4%) molluscum contagiosum, 9 (3.5%) candidiasis, 8 (3.1%) scabies, 6 (2.3%) squamous cell carcinomas, were included. The top-1 correct diagnosis rate of the application was 68.9%, compared to the 50.4% of the PCPs and 73.2% of the dermatologists. The application significantly outperformed PCPs with regard to the correlation with the gold standard diagnosis (P < 0.0001), and matched that of the dermatologists.

Conclusions: AI-based image diagnosis platforms can potentially be a low-cost rapid decision support tool for PCPs, integrated with syndromic management programs and direct-to-consumer services, and address healthcare inequities in managing genital dermatoses.