Development of an AI-based support system for controlled ovarian stimulation

Area of Science:

• Reproductive Medicine
• Artificial Intelligence in Healthcare
• Clinical Decision Support Systems

Background:

• Controlled ovarian stimulation (COS) is a critical component of in vitro fertilization (IVF) procedures.
• Optimizing COS protocols is essential for improving IVF success rates and patient outcomes.
• Current COS protocols often rely on physician expertise, which can lead to variability.

Purpose of the Study:

• To develop and validate an artificial intelligence (AI) system to support the implementation of COS protocols.
• To create AI models for predicting optimal oocyte retrieval timing and inferring medication prescriptions.
• To enhance the efficiency and consistency of COS in clinical practice.

Main Methods:

• Development of two AI models: an oocyte retrieval decision model and a prescription inference model.
• Training of AI models using a comprehensive dataset of IVF treatment records from a clinical management system.
• Evaluation of model performance using metrics such as area under the curve (AUC), sensitivity, and specificity.

Main Results:

• The oocyte retrieval decision model achieved a high AUC of 0.964.
• The prescription inference model demonstrated a strong performance with an AUC of 0.948.
• Individual prediction models within the system (hCG, hMG, Cetrorelix, Estradiol) showed AUC values ranging from 0.914 to 0.976.

Conclusions:

• The developed AI system demonstrates high accuracy and clinical utility in supporting COS protocols.
• The AI algorithm has been successfully implemented as a COS tool in a clinical setting.
• This AI-driven approach has the potential to standardize and improve COS practices in IVF treatments.