Integrating machine learning and SHAP interpretations as a decision-support tool for embryo quality assessment in...

Area of Science:

• Reproductive Biology
• Artificial Intelligence
• Animal Science

Background:

• Traditional embryo grading in cattle is subjective and lacks predictive power for pregnancy outcomes.
• Improving embryo quality assessment is vital for reproductive efficiency and genetic advancement in dairy herds.

Purpose of the Study:

• To develop and validate an explainable artificial intelligence (XAI) framework for predicting in vivo-derived embryo quality in dairy cattle.
• To provide transparent, biologically relevant insights into the factors influencing embryo quality prediction.

Main Methods:

• A dataset of 418 embryo records from Holstein cows was analyzed, incorporating 27 attributes including hormonal levels, follicle size, and body condition score.
• Six machine learning models were evaluated, with HistGradientBoost showing superior performance.
• SHAP (SHapley Additive exPlanations) was used for model interpretability, offering global and local feature explanations.

Main Results:

• The HistGradientBoost model achieved 71.1% accuracy and a 0.715 F1-score.
• Hormonal profiles (progesterone and estradiol) and accessory spermatozoa count were identified as dominant predictors of embryo quality.
• Local SHAP analysis highlighted specific hormonal markers, like post-GnRH progesterone, influencing individual embryo quality predictions.

Conclusions:

• The developed XAI framework accurately predicts dairy cattle embryo quality, offering a more objective assessment than traditional methods.
• Hormonal dynamics are critical determinants of embryo quality, providing valuable biological insights.
• This AI tool can support veterinarians in optimizing cattle breeding protocols and enhancing reproductive outcomes.