Innovations in the classification of embryos take the form of artificial intelligence systems that can evaluate embryos and accurately predict the results of introducing reproductive technologies. The Australian-American group of scientists on Wednesday presented a new study at the Scientific Congress and the exhibition ASRM 2018.
A team in the United States modeled a convolutional neural network to accurately predict the morphological quality of blastocysts based on fixed images.
50,392 images of 10,148 blastocysts were used in the study. The quality of the blastocysts corresponded to the categories of “good”, “satisfactory” and “poor”, which allowed us to present statistically different results of implantation. 18,000 of these images were used at the learning level of the initial IV algorithm, and then the remaining images were tested.
When recognizing the “bad” and “good” blastocysts, the algorithm was 97.52% accurate.
Meanwhile, the Australian team has developed an artificial intelligence system, which in 93% of cases correctly predicts that a particular embryo will develop before the fetus begins to have a heartbeat. The fully automated system analyzes temporary video sequences and does not require the use of the human factor, and, therefore, does not subject the embryo to undesirable changes in its behavior.
The task was to teach Artificial Intelligence to the skill of using temporal video sequences to analyze embryo development and forecasting by determining spatial temporal characteristics, regardless of the age of the mother and the ability of the embryo to demonstrate heartbeat. In the experiments, which lasted from 2014 to the present, eight laboratories from four countries of the world took part.
The 1 603 patients who participated in the study were in the age group from 22 to 50 years (that is, the average age was 35.6 years), while 10 208 embryos were included in the program, regardless of their stage of development or quality.
Commented on by Amy Sparks, President of SART: “Artificial intelligence applied to the analysis of the quality and potential of embryos has great prospects for improving the chances of patients for a successful pregnancy in a shorter time.”
Based on the materials of the ASRM Scientific Congress 2018