ARTIFICIAL INTELLIGENCE IN GYNECOLOGY AND WOMEN'S HEALTH: A NARRATIVE REVIEW OF CURRENT APPLICATIONS, CHALLENGES, AND FUTURE DIRECTIONS

Abstract

Background: Artificial intelligence (AI) is being used in more and more settings in obstetrics and gynecology, providing novel screening, diagnostic, treatment planning and monitoring solutions in women's health. AI tools can potentially address the gaps in service provision in low- and middle-income countries where medical access to specialist gynecologic and reproductive health care is not always available.

Objectives: To provide an update on the applications, benefits, limitations, and ethical implications of AI in gynecologic oncology, reproductive medicine, obstetric and gynecologic imaging, and gynecologic surgery.

Methods: A narrative literature review was conducted that incorporated structured data from PubMed, Scopus, and Google Scholar, and keywords included "artificial intelligence," "machine learning," "deep learning," "gynecology," "obstetrics," and "women's health," and combinations thereof, with the vast majority of searches occurring between 2020 and 2026. Key metrics were extracted from systematic reviews, meta-analyses and primary studies reporting quantitative diagnostic data (accuracy, sensitivity, specificity or area under the curve [AUC]).

Results: Twenty-one primary sources were included, representing over 150 individual primary sources. AI-enabled cervical cancer screening had an accuracy of up to 95% in 75 combined studies. The machine learning classifiers for PCOS diagnosis showed accuracy ranging from 89-100%, and AUC of 73-100% in 31 studies. The machine-learning model for the diagnosis of endometriosis had an AUC of 0.80 during validation. The pooled AUC of testing for the presence of PM in ovarian cancer was 0.81 for the radiomics-based models and 0.87 for the models that incorporated both clinical and radiomics-based data. Embryo-selection and IVF-outcome prediction models were more consistent than unaided visual grading, but prospective validation is still lacking.

Conclusion: AI has significant potential to enhance diagnostic precision, efficiency, and personalization of women's health care in the field of gynecology; however, its responsible use demands diverse training datasets, prospective validation, clear regulatory oversight, and careful efforts aimed at equity, particularly in low-resource countries like Pakistan.

Keywords: Artificial Intelligence, Machine Learning, Gynecology, Women's Health, Reproductive Medicine, Gynecologic Oncology, Narrative Review.