The following is translated from the original:
During in vitro fertilization (IVF), eggs and sperm produce many different embryos. The embryologist then selects the embryo with the greatest chance of successful pregnancy and transfers it to the patient.
Embryologists make this choice by using their expertise and applying a set of widely accepted principles based on the appearance of the embryo. In recent years, there has been a strong interest in using various artificial intelligence (AI) technologies in this process.
We developed one such artificial intelligence system and tested it in a study of more than 1000 IVF patients. Our system selects the same embryos as human experts in approximately two-thirds of cases, with only slightly lower overall success rates. The study results were published in the journal Nature Medicine.
Can deep learning help IVF?
Over the past few years, we have worked with colleagues in Sweden to develop software to identify which embryos are most likely to have successful in vitro fertilization. Our system uses deep learning, an artificial intelligence method of looking for patterns in large amounts of data.
While developing the system, we conducted a retrospective study to compare the choice of the system with real-world decisions that embryologists have made in the past. These early results suggest that deep learning systems may do better than human experts. Therefore, the next step is to correctly test the system through randomized trials.
Our trial involved 1,066 patients in 14 fertility clinics in Australia and Europe (Denmark, Sweden and the United Kingdom). For each patient, the deep learning system and human experts select which embryos to implant. Then, randomly choose which of the two to use.
This study is the first randomized controlled trial ever conducted on a deep learning system for embryo selection. Deep learning may have many medical applications, but this is one of the few prospective randomized trials of the technology in health care to date.
What we found?
In our research, we found that there is actually no difference between the two methods. When the deep learning system selects embryos, the clinical pregnancy rate (the likelihood of seeing the fetal heart after the first embryo transfer) is 46.5%, while when the embryologist selects embryos, the clinical pregnancy rate is 48.2%.
In other words, the difference is small. In fact, 65.8% of the time, the deep learning system selects the same embryo as the embryologist. However, we have also found that artificial intelligence systems complete embryo selection tasks ten times faster than embryologists.
One of the goals of our research is to prove the “non-inferiority” of our deep learning system. This is common in medical research because we always want to ensure that new technologies proposed do not lead to worse results than existing standards.
Although deep learning systems produce results that are very similar to those of human experts, our research does not completely clear the barriers to proving “non-inferiority.”
In fact, the overall success rate of this study was much higher than we expected. This changes the statistics, which means we need to conduct a larger study-involving nearly 8,000 patients-to prove that the new method is not inferior.
no significant difference
Previously, many ethical issues have been raised about deep learning in embryo selection. One is that biased selection of deep learning models may change the sex ratio, which ultimately produces more male or female embryos.
However, we found that deep learning embryo selection did not lead to a change in the sex ratio.
What we concluded from our research was that there was no significant difference in pregnancy rates between embryos selected by deep learning systems and embryos selected by experienced embryologists.
From this point of view, the use of deep learning tools for embryo selection will not fundamentally change the outcomes of patients undergoing in vitro fertilization (because it mostly selects the same embryos). However, using such reliable automated tools may make embryology laboratories more efficient and consistent.
The study further concluded that randomized trials that take years to conduct may not be the best way to study such rapidly developing technologies. Our future work to evaluate this technology will need to examine alternative but still clinically effective methods of the subject.
If you want to learn more, you can click on the link below the video.
Thank you for watching this video. If you like it, please subscribe and like it. thank
Original text:https://medicalxpress.com/news/2024-08-ai-ivf-embryos-human-randomized.html
More information: Peter J. Illingworth et al, Deep learning versus manual morphology-based embryo selection in IVF: a randomized, double-blind noninferiority trial, Nature Medicine (2024). DOI: 10.1038/s41591-024-03166-5
More information: Peter J. Illingworth et al., Deep learning and morphology-based artificial embryo selection in IVF: A randomized, double-blind non-inferiority trial, Nature Medicine (2024). DOI：10.1038/s41591-024-03166-5
Journal information: Nature Medicine
Journal information: Natural Medicine
Oil tubing: