Embryo through the lens: from time-lapse cinematography to artificial intelligence

Incidence, dynamics and recurrences of reverse cleavage in aneuploid, mosaic and euploid blastocysts, and its relationship with embryo quality

Background

During embryonic development, the normality of cleavage and the ploidy state are closely related to the final clinical outcome. At present, many research teams are focusing on the combined application of timelapse (TL) technology and preimplantation genetic testing (PGT) technology, hoping to find a connection between the two aspects of morphodynamics and genes. In the process of embryonic cleavage, there is a common abnormal cleavage pattern called reverse cleavage (RC). RC refers to blastomere fusion and failed cytokinesis. There are very few reports about it. Whether the occurrence of RC affects blastocyst euploidy is even less clear. Whether the RC phenomenon affects the embryonic developmental potential and whether it is related to the embryo ploidy. This is important for clinicians and embryologists. In this study, we used TL to observe whether there was a phenomenon of RC in each biopsy embryo and then combined it with the ploidy state to give an answer, which provided support for the selection strategy of RC embryos.

Methods

A total of 405 TL-PGT cycles and 1,467 blastocysts were included in the study. All TL data were collected from the Reproductive Medicine Center, Huazhong University of Science and Technology Hospital. Embryos images throughout embryonic development, from post-insemination to day 5 or 6 until biopsy and cryopreservation, were acquired by the Embryoscope Plus TL microscopy system from January 2019 to December 2020. This study investigated the overall incidence of RC during cleavage; the relationship between RC phenomenon and the number of occurrences and ploidy results; the relationship between RC occurrence and blastocyst developmental quality, as well as the dynamics of RC embryos.

Results

Among the 1,453 blastocysts biopsied, 400 blastocysts showed RC phenomenon at the cleavage stage, and the incidence rate was 25.9%. In euploid, mosaic and aneuploid embryos, the incidence of RC was 27.2%, 26.6%, and 25.0%, respectively. The incidence of RC was similar among these three groups with no significant difference (P > 0.05). The number of RC occurrences was not associated with embryo ploidy status (P > 0.05). In general, the blastocyst quality of the RC + group was lower than that of the RC- group. In the ICM score, the proportion of A score in the RC + group was significantly lower than that in RC- group (P < 0.05). In the TE score, there was no significant difference between the two groups of A-grade blastocysts, but the proportion of B-grade blastocysts in the RC + group was significantly lower than that in the RC- group (P < 0.01). In terms of developmental kinetic parameters, the cleavage synchrony parameters s2 and s3 were significantly longer in RC + embryos than in RC- embryos (P < 0.05). However, these changes in kinetic parameters were not significantly different between the euploid, mosaic and aneuploid groups.

Conclusions

The chromosomal euploidy of cleavage-stage embryos with RC phenomenon developed to the blastocyst stage was not significantly different from that of cleavage normal blastocysts. Therefore, RC embryos should not be discarded. It is recommended to select and utilize blastocyst culture, which has similar clinical value to normal cleavage embryos.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13048-022-01026-9.

Using deep learning to predict the outcome of live birth from more than 10,000 embryo data

Background

Recently, the combination of deep learning and time-lapse imaging provides an objective, standard and scientific solution for embryo selection. However, the reported studies were based on blastocyst formation or clinical pregnancy as the end point. To the best of our knowledge, there is no predictive model that uses the outcome of live birth as the predictive end point. Can a deep learning model predict the probability of live birth from time-lapse system?

Methods

This study retrospectively analyzed the time-lapse data and live birth outcomes of embryos samples from January 2018 to November 2019. We used the SGD optimizer with an initial learning rate of 0.025 and cosine learning rate reduction strategy. The network is randomly initialized and trained for 200 epochs from scratch. The model is quantitively evaluated over a hold-out test and a 5-fold cross-validation by the average area under the curve (AUC) of the receiver operating characteristic (ROC) curve.

Results

The deep learning model was able to predict live birth outcomes from time-lapse images with an AUC of 0.968 in 5-fold stratified cross-validation.

Conclusions

This research reported a deep learning model that predicts the live birth outcome of a single blastocyst transfer. This efficient model for predicting the outcome of live births can automatically analyze the time-lapse images of the patient’s embryos without the need for manual embryo annotation and evaluation, and then give a live birth prediction score for each embryo, and sort the embryos by the predicted value.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12884-021-04373-5.