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

Abstract

Time-lapse microscopy for embryos is a non-invasive technology used to characterize early embryo development. This study employs time-lapse microscopy and machine learning to elucidate changes in embryonic growth kinetics with maternal aging. We analyzed morphokinetic parameters of embryos from young and aged C57BL6/NJ mice via continuous imaging. Our findings show that aged embryos accelerated through cleavage stages (from 5-cells) to morula compared to younger counterparts, with no significant differences observed in later stages of blastulation. Unsupervised machine learning identified two distinct clusters comprising of embryos from aged or young donors. Moreover, in supervised learning, the extreme gradient boosting algorithm successfully predicted the age-related phenotype with 0.78 accuracy, 0.81 precision, and 0.83 recall following hyperparameter tuning. These results highlight two main scientific insights: maternal aging affects embryonic development pace, and artificial intelligence can differentiate between embryos from aged and young maternal mice by a non-invasive approach. Thus, machine learning can be used to identify morphokinetics phenotypes for further studies. This study has potential for future applications in selecting human embryos for embryo transfer, without or in complement with preimplantation genetic testing.

Keywords: machine learning; maternal aging; morphokinetics; predictive modeling; preimplantation mouse embryos; time-lapse microscopy.

Graphical Abstract

Figure 1

Schematic of methods. (A) Two cohorts of young (n = 5) and naturally aged (n = 10) female mice underwent ovarian hyperstimulation, oocyte harvest, and in vitro insemination, with time-lapse microscopy. (B) Following incubation of zygotes to the blastocyst stage, still images of each embryo were annotated manually and analyzed using statistical methods in the figure.

Figure 2

(A) Comparison of overall difference in the median number of hours for each morphokinetic time point between young (n = 15) and aged (n = 65) cohorts of embryos using two-way ANOVA (mixed-effect analysis of time vs age). Error bars represent upper and lower quartiles. (B) Further comparison of individual time points for each embryo using Kaplan–Meier survival estimates in the same experimental group. ns = non-significant, ^*P ≤ 0.05, ^**P ≤ 0.01, ^***P ≤ 0.001, ^****P ≤ 0.0001, 0 = young, and 1 = aged embryos.

Figure 3

Unsupervised clustering by Normal Mixtures. (A) Comparison of cluster performance by corrected AICc and the BIC. (B) Number of embryos in each cluster by normal mixtures clustering method. Normal mixtures model was used for unsupervised clustering due to overlapping distributions and its specificity for numerical data. Percent of aged or young embryos in each cluster. (C) Visual representation (biplot) of distribution of data for each cluster by PCs 1 and 4. Shaded area represents contour density with 90% confidence interval. Top oval with circular dots represents cluster 1 and thin oval with + signs represents cluster 2. Black circles surround the cluster center and their sizes are proportional to the count for each cluster.

Figure 4

(A) Diagram of supervised machine learning strategy and (B) performance of each model before and after hyperparameter tuning on the test set.