Clinical Report: AI models may improve PCOS detection

Overview

A systematic review of 80 studies indicates that AI models analyzing imaging, clinical, and molecular data may enhance the diagnosis and prediction of polycystic ovary syndrome (PCOS). These models have shown potential to outperform traditional diagnostic methods and identify molecular biomarkers associated with the condition.

Background

Polycystic ovary syndrome (PCOS) affects 8% to 13% of women of reproductive age globally and is often underdiagnosed due to variability in symptoms and diagnostic criteria. Traditional diagnostic methods, including clinical evaluation and imaging, have limitations that AI may help address. The integration of AI in PCOS management could lead to improved diagnostic accuracy and personalized treatment strategies.

Data Highlights

Key Findings

• AI models, including machine learning and deep learning, show diagnostic accuracies exceeding 95% in imaging studies.
• Supervised ML algorithms applied to clinical data consistently achieve high performance in predicting PCOS.
• Candidate biomarkers for risk stratification identified include HDDC3, SDC2, MAP1LC3A, and OVGP1.
• About 25% of studies utilized explainable AI techniques, enhancing interpretability.
• Key limitations include small sample sizes, class imbalance, and lack of external validation.
• LLMs like ChatGPT are being explored for clinical applications, but concerns about accuracy and reliability persist.

Clinical Implications

The findings suggest that AI could significantly improve the early diagnosis and management of PCOS, potentially leading to better patient outcomes. However, clinicians should remain cautious about the limitations of current AI models, including issues of interpretability and validation.

Conclusion

AI has the potential to transform PCOS diagnosis and management, but successful integration into clinical practice will require addressing current limitations and ensuring clinician trust in these technologies.

References

1. Ghaderzadeh M, et al., BMC Medical Informatics and Decision Making, 2025 -- Artificial intelligence in polycystic ovary syndrome: a systematic review of diagnostic and predictive applications
2. The Journal of Clinical Endocrinology & Metabolism, 2023 -- Evaluating Patients: Navigating Diagnostic Obstacles in Polycystic Ovary Syndrome Assessment
3. The Journal of Clinical Endocrinology & Metabolism, 2023 -- The Quest for Identifying Polycystic Ovary Syndrome
4. The asco post, 2025 -- AI May Improve Ovarian Cancer Diagnoses
5. American Journal of Epidemiology — Assessment of the Frequency and Associated Factors of Diagnosed and Likely Polycystic Ovary Syndrome (PCOS) in a Sample of Parous Women
6. https://academic.oup.com/jcem/article/108/10/2447/7242360
7. Artificial intelligence in polycystic ovary syndrome: a systematic review of diagnostic and predictive applications | BMC Medical Informatics and Decision Making | Springer Nature Link