Machine learning and microfluidic integration for oocyte quality prediction.

Area of Science:

• Biomedical Engineering
• Reproductive Technology
• Machine Learning

Background:

• In vitro fertilization (IVF) success rates remain suboptimal, highlighting a need for improved methods in oocyte quality assessment.
• Current methods for evaluating oocyte quality often lack objectivity and predictive accuracy.

Purpose of the Study:

• To develop and evaluate a microfluidic-based machine learning framework for enhanced oocyte quality prediction.
• To integrate biomechanical features with machine learning algorithms to improve the accuracy of predicting IVF outcomes.

Main Methods:

• Immature oocytes were analyzed in a microfluidic channel, extracting biomechanical features like Cortical Tension (CT) and Deformation Index (DI).
• Oocyte diameter and critical flow rate (Q) were also measured.
• Supervised (e.g., Random Forest) and unsupervised (e.g., Agglomerative Clustering) machine learning models were applied to a dataset of 54 oocytes.

Main Results:

• The Random Forest model achieved the highest classification accuracy (76.10% with K-Fold cross-validation).
• Agglomerative Clustering demonstrated effective grouping patterns among oocytes (Silhouette score = 0.49).
• The study successfully integrated biomechanical profiling with machine learning for objective oocyte assessment.

Conclusions:

• The proposed microfluidic-machine learning framework significantly enhances the objectivity and accuracy of oocyte quality prediction.
• This approach shows promise for improving embryo selection strategies and optimizing IVF outcomes in Assisted Reproductive Technology (ART).