Genetic algorithm-assisted machine learning for clinical pregnancy prediction in in vitro fertilization

Area of Science:

• Reproductive medicine
• Artificial intelligence in healthcare
• Medical data analysis

Background:

• Developed a clinical pregnancy prediction model using machine learning.
• Integrated static embryo images and clinical data for in vitro fertilization (IVF) outcome prediction.

Purpose of the Study:

• To create a decision-support system for critical IVF cycle choices.
• To enhance embryo selection accuracy through predictive modeling.

Main Methods:

• Collected historical data from 697 IVF patients, including embryo images.
• Applied and compared machine learning algorithms: decision tree, random forest, and gradient boosting.
• Optimized algorithms using a genetic algorithm for improved performance.

Main Results:

• Achieved a peak prediction accuracy of approximately 65%.
• Gradient boosting model demonstrated superior performance in predicting clinical pregnancy.
• Performance differences among algorithms were statistically significant based on various metrics.

Conclusions:

• This study represents a foundational step for ML-based IVF prediction models.
• Further validation is necessary to enhance model performance for clinical application.
• The findings suggest potential for AI to improve IVF decision-making.