Machine learning in time-lapse imaging to differentiate embryos from young vs old mice†

Area of Science:

• Developmental Biology
• Reproductive Medicine
• Artificial Intelligence in Medicine

Background:

• Embryonic development kinetics are crucial for reproductive success.
• Maternal aging is associated with decreased fertility and increased risk of developmental abnormalities.
• Non-invasive methods for assessing embryo quality are highly desirable.

Purpose of the Study:

• To investigate the impact of maternal aging on embryonic growth kinetics using time-lapse microscopy.
• To apply machine learning algorithms to differentiate embryos based on maternal age.
• To explore the potential of AI in identifying age-related embryonic phenotypes.

Main Methods:

• Continuous time-lapse imaging of mouse embryos (C57BL6/NJ) from young and aged donors.
• Analysis of key morphokinetic parameters during early development.
• Application of unsupervised and supervised machine learning (Extreme Gradient Boosting) for classification.

Main Results:

• Embryos from aged mothers exhibited accelerated progression through cleavage to morula stages.
• No significant differences in blastulation stage timing were observed between age groups.
• Machine learning models accurately predicted maternal age-related phenotypes (Accuracy: 0.78, Precision: 0.81, Recall: 0.83).

Conclusions:

• Maternal aging alters the pace of embryonic development, particularly in early cleavage stages.
• Artificial intelligence, via machine learning, can non-invasively identify age-dependent embryonic morphokinetic signatures.
• This approach holds promise for improving human embryo selection for assisted reproductive technologies.