Optimizing predictive features using machine learning for early miscarriage risk following single vitrified-warmed...

Area of Science:

• Reproductive medicine
• Artificial intelligence in healthcare
• Biostatistics

Background:

• Early miscarriage following single vitrified-warmed blastocyst transfer (SVBT) poses a significant challenge in assisted reproductive technology (ART).
• Accurate prediction of miscarriage risk is crucial for effective patient counseling and clinical decision-making.

Purpose of the Study:

• To evaluate the efficacy of machine learning (ML) models in predicting the risk of early miscarriage after SVBT.
• To compare the performance of various ML models against traditional statistical methods.

Main Methods:

• A retrospective analysis of 1,664 SVBT cycles from two centers was performed.
• Multiple ML models, including Logistic Regression, Random Forest, Gradient Boosting, and Voting Classifier, were developed and evaluated.
• Key predictors were identified using Mutual Information and Recursive Feature Elimination (RFE).

Main Results:

• Maternal age, paternal age, endometrial thickness, blastocyst quality, and ovarian stimulation parameters were identified as critical predictors.
• Ensemble ML models significantly outperformed logistic regression (AUC=0.584).
• The Voting Classifier achieved the highest AUC (0.836), accuracy (0.780), precision (0.914), and specificity (0.942).

Conclusions:

• Ensemble ML models, particularly Voting Classifier and Gradient Boosting Classifier, offer superior prediction of early miscarriage risk post-SVBT.
• These advanced models enable accurate, individualized risk assessments, improving clinical decisions in ART.
• The findings support the integration of ML for personalized patient care in reproductive medicine.