Author:
Shurygina O. V.,Nemkovskiy G. B.,Rusakov D. Y.,Gromenko D. S.,Taxants M. I.,Novikova E. V.,Vasilenko O. Y.,Tugushev M.T.,Shipulin N. A.,Kuznetsov A. B.,Belyakov V. K.
Abstract
Relevance: Currently, it is extremely important to identify predictors of the development of a competent embryo that determine its implantation potential. In this case, the predictors are predictive parameters that should be assessed together to rank and select human embryos.
We introduced the concept of «human embryo morphodynamic profile» to standardize the description of the development of human embryos cultured in vitro. We identified a set of morphokinetic states that are included in the profile and located on the time scale depending on the moment of their registration. All timing cutoffs (points) are given in chronological order relative to the moment of fertilization.
The purpose of the study was to implement an information system utilizing artificial intelligence technologies for an automated formation of the morphodynamic profile of a human embryo based on time-lapse photography of the process of human embryo cultivating to the blastocyst stage.
Materials and methods: Visual information about the pre-implantation development of human embryos to the blastocyst stage (0 - 6 days from insemination) was collected using an «Embryovisor» incubator for IVF laboratories with a time-lapse (hyperlapse) video fixation system (LLC «WESTTRADE LTD,” Russia). The embryos were cultivated individually in special microwells of WOW dishes (Vitrolife, Sweden). Visual information about cultured human embryos was collected, marked, and prepared at the Laboratory of assisted reproductive technologies (ART) of the Clinical Hospital IDK CJSC “Medical Company IDK” (Group of Companies “Mother and Child,” Samara, Russia) and the medical center “Semya” (Ufa, Russia). The morphodynamic profile was marked using the EmbryoVisor software (customized version). Graphics and markup information was uploaded to the SberCloud cluster. A convolutional neural network for solving the multiclass classification task was implemented on the Christofari supercomputer of the SberCloud cluster.
Results: Based on the available database, we have developed a system for forming the morphodynamic profile of a human embryo, taking into account the placement of markers of fixed morphokinetic states.
Conclusion: The ability to record major morphodynamic events and assess them allows a more comprehensive approach to evaluating and ranking developing embryos and selecting the most promising embryo for implantation.
Reference12 articles.
1. Meseguer M. Fertil. Steril., 2016; 105(2): 295-296. https://doi.org/10.1016/j.fertnstert.2015.12.126;
2. Stensen M.H., Tanbo T.G., Storeng R., Abyholm T., Fedorcsak P. Fertil. Steril., 2015; 103(2): 374-381. https://doi.org/10.1016/j.fertnstert.2014.10.031;
3. Ren S., He K., Girshick R., Sun J. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137- 1149. https://doi.org/10.1109/TPAMI.2016.2577031;
4. He K., Zhang X., Ren S., Sun J. Deep Residual Learning for Image Recognition // 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), December 2016. https://doi.org/10.1109/CVPR.2016.90;
5. Tan M., Le Q.V. EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks // International Conference on Machine Learning, May 2019, https://arxiv.org/abs/1905.11946;
Cited by
1 articles.
订阅此论文施引文献
订阅此论文施引文献，注册后可以免费订阅5篇论文的施引文献，订阅后可以查看论文全部施引文献