AI In Personalized Treatment Plans For Endometriosis A Data-Driven Approach To Therapy Selection

• Published in: Frontiers in health informatics pp. 895 - 901
• Main Authors: Latif, Nayar, Babe, Rubina, Raz, Kainat, Zama, Uzma, Sherbano, Sherbano, Zaman, Faiza
• Format: Journal Article
• Language: English
• Published: 18.06.2025
• ISSN: 2676-7104; 2676-7104
• DOI: 10.63682/fhi2621

Background: Endometriosis is a chronic inflammatory disease that affects around 10% of women of reproductive age. It is a cause of pelvic pain, infertility and reduced quality of life. Varying efficacy and patient dissatisfaction is a result of current treatment strategies being empirical and not patient-oriented. Data Integration and Predictive Modelling: AI-based solutions may aid precise therapy selection through data integration and predictive modeling. Objectives:To assess the accuracy of AI algorithms to discover personalized therapeutic approaches for endometriosis from clinical, genetic and imaging data. Study design: A Retrospective Study. Place and duration of study: Department of  Gynae Gomal Medical College Dera Ismail Khan Pakistan from January to dec 2023 Methods:100 endometriosis patients from Department of  Gynae Gomal Medical College Dera Ismail Khan Pakistan from January to dec 2023. The machine learning models (random forest, logistic regression, Boost) were trained using patient features including the hormonal receptor status, the imaging findings, the genetic polymorphisms, and the prior treatment outcomes. The data was split into training (70%) and validation (30%) sets. We evaluated the performance of the model using AUC-ROC, F1-score, precision and recall metrics for the 10%, 20%, 30% and baseline data set. Results:100 patients were analyzed, with a mean age of 32.7 ± 6.4 years. AUC-ROC with 0.91 proved Boost algorithm to be the runaway winner. Patients chosen for AI-guided therapy also showed a 27 percent greater rate of symptom resolution over those in usual care (p = 0.002). There was also an 18% decrease in treatment switching (p = 0.01), suggesting better compatibility of initial therapy. Predictive features were ESR1 and CYP2C19 polymorphisms, location of lesions, and treatment history. Conclusion:AI-based models predicted the best treatment for endometriosis with high accuracy, ensuring better clinical outcomes and fewer failed trial-and-error approaches. Together, these findings bolster the use of AI as part of personalized care pathways. Additional prospective studies are needed for clinical application. Warning: The above citation should be considered while referencing this article.