Use of AI in Identification of Sexually Transmitted Infections and Anogenital Dermatoses: A Systematic Review and Meta-Analysis.

Abstract

Importance: Artificial intelligence (AI) excels in dermatology. However, its applications to sexually transmitted infections (STIs) remain unclear.

Objective: To assess the performance of AI algorithms and their applications in detecting STIs and anogenital dermatoses from clinical images in sexual health.

Data sources: Six databases (IEEE Xplore, Embase, Scopus, Medline, Web of Science, and CINAHL) were searched for studies published from January 1, 2010, to April 12, 2024, using 3 main concepts: artificial intelligence, diagnosis, and sexually transmitted infections.

Study selection: Studies that used AI to identify anogenital skin conditions from clinical images were included. Studies that used non-AI approaches or nonanogenital conditions, as well as reviews and studies lacking performance metrics, were excluded.

Data extraction and synthesis: Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, 2 reviewers independently assessed full-text articles and extracted data using a standardized spreadsheet. Another 2 reviewers resolved any disagreements. A modified Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) critical appraisal tool and the Checklist for Evaluation of Image-Based AI Reports in Dermatology (CLEAR Derm) were used for quality assessment.

Main outcomes and measures: Pooled sensitivity and specificity of AI applications for detecting anogenital skin conditions. A bivariate random-effects meta-analysis was conducted for conditions with more than 3 studies.

Results: Of 5381 studies screened and 258 full texts selected, 140 met the inclusion criteria. Most studies reported on mpox (110 [78.6%]), while other anogenital conditions, including genital herpes (7 [5.0%]), genital warts (8 [5.7%]), scabies (8 [5.7%]), and molluscum contagiosum (6 [4.3%]), received less attention. Meta-analyses showed high performance of AI for identification of mpox (pooled sensitivity: 0.96 [95% CI, 0.93-0.97]; pooled specificity: 0.98 [95% CI, 0.97-0.99]), herpes simplex (sensitivity: 0.91 [95% CI, 0.71-0.98]; specificity: 0.97 [95% CI, 0.94-0.98]), genital warts (sensitivity: 0.87 [95% CI, 0.67-0.96]; specificity: 0.98 [95% CI, 0.95-0.99]), psoriasis (sensitivity: 0.90 [95% CI, 0.78-0.95]; specificity: 0.98 [95% CI, 0.96-0.99]), and scabies (sensitivity: 0.89 [95% CI, 0.84-0.93]; specificity: 0.98 [95% CI, 0.95-0.99]). Study quality was variable, and the assessment identified high risk of bias across the population selection (76.1%), reference standards (76.1%), and index tests (20.0%). Most studies relied on open-source datasets (121 [86.4%]); only 17 (12.1%) used external validation. All but 1 study (0.7%) remained at the proof-of-concept stage, and models were not publicly available for external evaluation.

Conclusions and relevance: The findings suggest that AI shows promise in identifying STIs and anogenital dermatoses but that significant research gaps exist. Future work should prioritize understudied STIs and differential conditions while improving data quality, conducting external validation, and validating findings in clinical settings.