Comparative study of machine learning approaches integrated with genetic algorithm for IVF success prediction.

Area of Science:

• Reproductive Medicine
• Artificial Intelligence
• Biomedical Data Science

Background:

• In Vitro Fertilization (IVF) success rates remain around 30%, necessitating improvements due to significant patient impact.
• Enhancing IVF outcomes is critical for addressing the emotional, financial, and health burdens faced by infertile couples.

Purpose of the Study:

• To develop and compare machine learning models for predicting IVF success.
• To identify key predictors of successful IVF outcomes.

Main Methods:

• Evaluated five machine learning algorithms: Random Forest, Artificial Neural Network (ANN), Support Vector Machine (SVM), Recursive Partitioning and Regression Trees (RPART), and AdaBoost.
• Utilized Genetic Algorithm (GA) for feature selection to enhance model performance and identify crucial predictive factors.

Main Results:

• AdaBoost combined with GA achieved the highest prediction accuracy at 89.8%.
• Random Forest with GA also showed strong performance (87.4% accuracy).
• GA significantly improved all tested classifiers, highlighting the importance of feature selection. Identified ten key predictors including female age, AMH, endometrial thickness, sperm count, and oocyte/embryo quality indicators.

Conclusions:

• Machine learning models, especially with feature selection, show significant potential for improving IVF outcome prediction.
• These predictive capabilities can assist clinicians in developing personalized treatment strategies for IVF patients.
• Further research and clinical validation are recommended to integrate these models into standard IVF practice.