Artificial intelligence learns to identify good embryos for IVF
Researchers have trained artificial intelligence software to grade human embryos with a high degree of accuracy.
In a study published in NPJ Digital Medicine, a multidisciplinary team led by researchers at Cornell University in New York City, New York, demonstrated the use of a Google deep-learning network to grade five-day-old embryos.
The algorithm, called ‘STORK’, was trained using time-lapse recordings of the development of over 10,000 embryos.
Embryo grading is a standard tool for embryo selection in IVF – with the embryos that score highly for factors such as symmetry, high cell number and low fragmentation being used for transfer first. However, these morphological assessments are time-consuming and subjective, with different embryologists coming to different conclusions. Embryo-grading schemes lack standardisation across the field. AI may be able to overcome these limitations.
‘It was an opportunity to automate a process that is time-consuming and prone to errors, something that’s not really been done before with human embryos,’
said Dr Olivier Elemento, study author and director of Weill Cornell’s Englander Institute for Precision Medicine.
The grading of embryos as poor, fair or high quality was compared between STORK and five embryologists from three different continents. STORK was demonstrated to be in line with the majority decision of the embryologists (3 out of 5) in 95.7 percent of cases. The embryologists themselves were in full agreement in only 89 of 374 embryos.
The group proposes that the tool will aid in boosting pregnancy rates and supporting single-embryo transfer, thus reducing the risks associated with multiple births.
‘Our algorithm will help embryologists maximise the chances that their patients will have a single healthy pregnancy,’ said Dr Elemento. ‘The IVF procedure will remain the same, but we’ll be able to improve outcomes by harnessing the power of artificial intelligence.’
Combining STORK scoring for 2182 embryos with patient information such as age and clinical diagnosis, the framework was used to predict scenarios with the likelihood of pregnancy. However, this would require further independent tests to demonstrate a clinical benefit.
‘All this algorithm can do is change the order of which embryos we transfer. It needs more evidence to say it helps women get pregnant quicker and safer.’
The group is planning a randomised control trial to compare implantation rates between STORK and embryologist-selected embryos. Further, they are reportedly working on using the system to identify embryos with aneuploidy, abnormal chromosome content. This could represent an alternative for preimplantation genetic screening, an expensive and invasive procedure.
‘We think the patterns of cell division we can capture with these movies could potentially carry information about these defects, which are hidden in just the snapshots,’
said Dr Elemento.