Machine-learning models evaluate embryo morphokinetics to detect subclinical signs of heat stress in embryos produced by multiple ovulation embryo transfer

Abstract

Increased demand for animal-based protein, driven by the in­creasing world population, has emphasized the use of bovine embryo transfer (ET). However, this demand is predominantly in locations where environmental conditions, like heat stress, are inhibiting reproductive performance. Heat stressed do­nors have decreased viable blastocyst production, greater vari­ability in embryonic gene expression and lower ET pregnancy outcomes in current literature. These embryonic changes are nonidentifiable to the human eye, highlighting the need for solutions to identify compromised embryos and improve ET efficiencies. Machine-learning artificial intelligence has been used to evaluate health by assessing embryo morphokinetics, or time-specific morphological changes during embryo devel­opment, from videos of embryos in standard media. Thus, the objective was to use machine learning to detect real-time mor­phokinetic activity of embryos from donor cows undergoing multiple ovulation embryo transfer (MOET) based on seasonal exposure to heat stress.