Artificial Intelligence in Endometriosis Management: A Guideline-Concordance Study

Abstract

Objective: This study aimed to evaluate how closely artificial intelligence (AI)- generated clinical recommendations correspond with established international guidelines for endometriosis management, using structured clinical scenarios.Methods: This study assessed the consistency of AI-generated responses with the 2022 ESHRE Endometriosis Guideline. Fifteen predefined clinical scenarios were designed to reflect the main areas of endometriosis management. Each scenario was submitted using a standardized prompt to ensure consistency. AI responses were evaluated using predefined guideline-based assessment matrices to determine concordance and to explore performance differences across clinical domains.Results: A total of 45 AI-generated responses across 15 clinical scenarios were evaluated. Overall concordance with ESHRE guideline-based reference answers was 93.3% (42/45). Diagnostic scenarios demonstrated 88.9% concordance, while medical management, surgical/multidisciplinary, and high-risk scenarios demonstrated 100% concordance. Fertility-related scenarios showed 83.3% concordance. Complete inter-run consistency was observed in 13 of 15 scenarios (86.7%), with variability limited to Cases 2 and 3. Fleiss’ kappa indicated a high level of inter-run agreement within the evaluated scenarios (κ = 0.831; 95% CI, 0.583–1.000).Conclusion: Artificial intelligence systems show meaningful alignment with evidence-based guidance in fundamental aspects of endometriosis management. However, variability in complex scenarios reveals important limitations. Although AI tools may support education or serve as an adjunct in clinical decision-making, they cannot replace expert clinical judgment in specialized gynecologic care.