Enhancing frozen-thawed embryo transfer outcomes and treatment personalization through machine learning models

Area of Science:

• Reproductive Medicine
• Artificial Intelligence in Healthcare
• Biostatistics

Background:

• Infertility impacts millions globally, with assisted reproductive technology (ART) crucial for treatment.
• Frozen-thawed embryo transfer (FET) success rates vary significantly (29.6-59.2%), necessitating improved prediction and personalization.
• Current ART outcomes highlight a need for advanced predictive analytics to enhance clinical pregnancy rates.

Purpose of the Study:

• To develop and validate machine learning (ML) models for predicting clinical pregnancy outcomes in FET cycles.
• To simulate personalized treatment strategies using ML models based on individual patient profiles.
• To enhance the precision and success rates of assisted reproductive technology through data-driven insights.

Main Methods:

• Retrospective analysis of 1013 FET cycles from two medical centers.
• Training and evaluation of four ML models (XGBoost, random forest, logistic regression, deep neural networks) using diverse feature sets.
• Performance assessment via ROC AUC, sensitivity, specificity; SHAP and decision curve analyses for predictor identification and clinical utility.

Main Results:

• The XGBoost model, utilizing combined and expert-selected clinical features, achieved the highest predictive performance (ROC AUC: 0.7922).
• Key predictors identified by SHAP analysis included embryo quality, female age, and anti-Müllerian hormone levels.
• Decision curve analysis confirmed the clinical utility of the XGBoost model, demonstrating its effectiveness in optimizing treatment decisions.

Conclusions:

• XGBoost-based ML models offer a robust framework for predicting FET outcomes and personalizing patient treatment.
• Integration of clinical and embryological factors enables precision care strategies, optimizing success rates in reproductive medicine.
• This study demonstrates the transformative potential of ML in advancing reproductive medicine and reducing the global burden of infertility.