An intelligent decision-making system for embryo transfer in reproductive technology: a machine learning-based...

Area of Science:

• Reproductive Medicine
• Artificial Intelligence in Healthcare
• Biostatistics

Background:

• Infertility affects public health, with assisted reproductive technology (ART) as a primary treatment.
• Current ART methods face limitations due to high costs and physical discomfort.
• Predicting optimal embryo transfer numbers is crucial for ART success.

Purpose of the Study:

• To develop machine learning (ML) decision-support models for predicting optimal embryo transfer numbers in ART.
• To classify cases into transferring two or fewer versus three or four embryos.

Main Methods:

• Utilized data from infertile couples via literature reviews.
• Developed binary classification models using Random Forest (RF), Logistic Regression (LR), Support Vector Machine (SVM), and Artificial Neural Network (ANN).
• Considered seven criteria: woman's age, sperm origin, embryo developmental qualities, infertility duration, woman's assessment, embryo morphology, and oocyte count.

Main Results:

• SVM model achieved the highest average accuracy (95.83%), followed by ANN and LR (91.67%).
• RF model showed the lowest variability with 88.89% accuracy.
• On new datasets, ANN and SVM models achieved 100% accuracy; RF and LR achieved 91.68%.

Conclusions:

• ML models, particularly SVM and ANN, effectively predict optimal embryo numbers for ART.
• These models demonstrate superior generalization and can guide ART clinical decisions.
• Enhanced decision support can potentially improve ART outcomes and patient experience.