In recent times, the development and implementation of Artificial Intelligence (AI) has demonstrated its potential to address inefficiencies in different stages of assisted reproduction. Among them, one of the most crucial processes that AI has shown potential to improve takes place in the in vitro fertilisation (IVF) laboratory: embryo selection.
Prompted by this potential, IVI set out to evaluate what AI can do for embryo selection by carrying out the most extensive study in the area of AI and fertility treatment in scientific history.
Applying AI to embryo selection
Our recent study analysed over 25,000 embryos and 4,000 patients, making it the largest of its kind in a clinical research setting. This has given way to a new approach to embryo selection that is universal, standardised and automatic, and can be applied in IVF laboratories in IVI clinics worldwide. Marked by its potential to revolutionise the embryology sector, the research has been published in the North American journal Fertility and Sterility and European Reproductive Biology OL.
“In our embryology laboratories, we applied solutions based on data, which allowed us to evaluate the potential for embryo implantation with new levels of accuracy. This, in turn, serves to improve the efficiency of one of the most important processes in assisted reproduction: embryo culture and selection. We have seen accuracy levels of 75% in the selection of chromosomally normal embryos. Meanwhile, with traditional processes of manual evaluation, it would not have been possible to identify these embryos with this level of accuracy, regardless of the experience of the embryologist,” explains Dr. Marcos Meseguer, embryologist and scientific supervisor of the Embryology Unit of IVI Valencia. Recently, Dr. Meseguer was recognised by US instituion Stanford University as one of the world’s best researchers, alongside Professors José Remohí, Antonio Pellicer and Dr. Juan Antonio García Velasco, all of whom belong to IVI.
Results presented at the 37th Annual Congress of ESHRE
The research was presented on the 28th June 2021 at the 37th Congress of the European Society for Human Reproduction and Embryology (ESHRE) under the title: “Computer vision can distinguish between euploid and aneuploid embryos. A novel artificial intelligence (AI) approach to measure cell division activity associated with chromosomal status”. The event – which remains among the most important in the sector – was once again held virtually the second consecutive year.
Doctoral student Lorena Bori, based at IVI Valencia, presented the key findings of the study, which was co-directed by Dr. Meseguer and Dr. Daniella Gilboa from Tel-Aviv.
- The purpose of the study was to identify a chromosomally normal embryo – (“euploid”) – without having to apply invasive techniques, such as the extraction of cells from the blastocyst applied during an embryo biopsy procedure.
- For the first time, an AI-based system is able to accurately analyse the initial stages of embryonic development. While quantifying the duration of cell cycles and determining the diameter of the cells that form the blastocyst, an intelligent algorithm is generated capable of distinguishing between a chromosomally normal and abnormal embryo with an accuracy level of 75%.
- During the study, a record 25,000 embryos were genetically analysed, which makes it the largest study ever performed at a scientific level. This research confirms, with a high level of statistical accuracy, that embryos behave differently in their pattern of development and behaviour depending on their chromosomal content, and that this can be applied automatically by combining AI and time-lapse imaging technology.
- The results show that euploid embryos begin their development into blastocysts earlier than aneuploid embryos. This longer period for aneuploid embryos to grow until they reach the blastocyst phase is explained by their higher level of cellular activity.
- This study could bring revolutionary advances in the process of categorising and selecting embryos at a chromosomal level, which could lead to increased implantation and pregnancy rates. Furthermore, by reducing the chances of transferring embryos with chromosomal abnormalities, this fast, non-invasive and inexpensive technique could significantly increase live birth rates, particularly for specific patient groups where there is an increased risk.
- When applied, this technique could also revolutionise current processes in assisted reproduction by bypassing the use of invasive methods that, in part, can affect the viability of the embryo. The current results are within the same range found with invasive pre-implantation genetic testing for aneuploidy (PGT-A), yet remove the financial cost to patients and potential damage to the embryo. In addition, it would streamline workflows within embryology laboratories by automating a process currently produced manually, and by hand.
Based on these results, IVI can affirm that this completely non-invasive technique would improve all current methods of embryo selection.
How can AI improve IVF treatments?
AI is a broad term that encompasses both machine learning and deep learning. It refers to any program capable of learning from experience to solve problems and performing tasks as human beings normally do.
“Essentially, this AI-based system classifies embryos automatically through methods based on the experience of entire teams of expert embryologists. It detects and evaluates all the development steps of the embryo while also classifying its morphology, all in real time. By automating the process of embryo selection based on this data, it is more accurate when compared to manual assessment and selection. For this reason, the likelihood of an on-going pregnancy is directly related to the percentage score, which allows the patient to have a higher chance of success,” says Dr. Meseguer.
- Analyse big data to offer broader and earlier identification of new indicators of infertility and correlations in behavior patterns.
- Provide more accurate predications of IVF success and help to manage patient expectations in terms of treatment outcomes and live births.
- Increase success rates by allowing for more personalised and individualised treatment protocols.
- Use intelligent algorithms and deep learning to support clinical decision-making throughout all laboratory procedures, such as identifying the most viable gametes and embryos.
- Calculate the reception status of the endometrium using intelligent algorithms and unique biomarkers to determine the optimal time to transfer an embryo.
For more information, please contact Press.UK@ivirma.com.