To propose AutoPCOS, an innovative multimodal intelligent framework for PCOS risk stratification and diagnostic support.
Key Findings:
The framework showed robust predictive performance across different data availability scenarios, with precision values exceeding 0.929 in specific subgroups.
Models achieved precision values ≥ 0.929 in subgroups with BMI < 24 and irregular menstrual cycles, indicating high accuracy.
Random Forest outperformed other models in comparative analysis, demonstrating its effectiveness as a primary classifier.
Integration of a knowledge base and Lingshu large language model provided interpretable risk explanations, enhancing user understanding.
Interpretation:
AutoPCOS offers a flexible and resource-aware approach for PCOS risk assessment, significantly enhancing decision-making and interpretability for both patients and healthcare providers.
Limitations:
Future validation is needed using diverse real-world clinical datasets to ensure applicability.
Model generalizability requires further improvement to adapt to various clinical settings.
Conclusion:
AutoPCOS has the potential to serve as a practical tool for PCOS risk assessment and diagnosis, particularly in resource-limited settings, addressing current diagnostic challenges.
