Selection of human embryo for IVF treatment using ensemble machine learning technique

Area of Science:

• Reproductive Medicine
• Artificial Intelligence in Healthcare
• Medical Imaging Analysis

Background:

• Embryo selection is crucial for in vitro fertilization (IVF) success but is subjective and relies on embryologist expertise.
• Current methods for embryo assessment lack objectivity and consistency, potentially impacting IVF outcomes.

Purpose of the Study:

• To develop a comprehensive, AI-driven framework for objective embryo quality assessment.
• To enhance the accuracy and reliability of embryo selection in IVF procedures.

Main Methods:

• Image pre-processing of day 3 embryos and blastocysts.
• Extraction of multifaceted image features using local descriptors.
• Feature selection via Extra Trees classifier and deep feature extraction using a 1D-Convolutional Neural Network (1D-CNN).
• Ensemble of classifiers replacing the 1D-CNN's final layer for robust embryo quality determination.

Main Results:

• The proposed AI framework achieved high accuracy rates: 93% for blastocyst and 98% for day 3 embryo datasets.
• The 1D-CNN enhanced feature interpretability and predictive power.
• The ensemble classifier approach provided a robust decision-making framework.

Conclusions:

• The developed AI framework offers a significant improvement over existing methods for embryo quality assessment.
• This objective approach has the potential to increase IVF treatment success rates.
• The study highlights the efficacy of combining advanced image analysis with machine learning for reproductive medicine applications.