AI tools help choose best embryos for IVF
Artificial intelligence (AI) is increasingly being used to assist in selecting the most promising embryos for in vitro fertilisation (IVF), raising both excitement and ethical questions. IVF, a procedure that allows millions of children to be born each year to parents struggling with infertility, involves fertilising an egg with sperm in a laboratory. However, success rates can vary widely and tend to decline with maternal age.
Nearly 50 years after the birth of the first IVF child, AI is emerging as a tool to help clinicians identify embryos with the highest potential for implantation, Nathalie Massin, head of the clinical unit at the American Hospital of Paris’ IVF centre, told AFP. The hospital performs more than 2,300 IVF procedures annually and uses an embryoscope—a time-lapse camera that continuously records embryo development.
Previously, the information captured by the embryoscope, including embryo shape, symmetry, and cell division patterns, was only partially utilised. Now, AI tools, particularly those employing machine learning, can analyse this data to help doctors select embryos most likely to result in successful pregnancies or suitable for freezing. This reduces the need for multiple, costly IVF attempts by identifying embryos that may have abnormalities associated with miscarriage. Importantly, AI achieves this without manipulating embryos, addressing some concerns about “designer babies.”
Frida Entezami, co-leader of the hospital’s IVF department, emphasised that human doctors will remain in control of final decisions, with AI serving as an additional tool. The hospital recently adopted an AI system from Israeli start-up AIVF, currently being trialled internally with the goal of halving the number of cycles needed to achieve pregnancy. According to Entezami, AIVF can offer a 70% probability that recommended embryos are free of genetic abnormalities—a notable improvement, given that roughly half of pre-implantation embryos currently exhibit such issues.
Beyond embryo selection, AI can optimise other aspects of IVF. Algorithms can help adjust hormone injection timing and dosage before egg retrieval and improve sperm selection, particularly in samples with low sperm counts. Anne-Claire Lepretre, head of France’s Biomedicine Agency’s assisted reproductive technologies (ART) unit, highlighted that AI models can learn from previous failed attempts, gradually increasing the likelihood of success for future cycles. This personalised approach has the potential to ease the emotional burden of the often long, complex, and psychologically taxing IVF process.
Despite these benefits, some experts caution against uncritical adoption. Julian Koplin, a bioethicist at Australia’s Monash University, noted that using AI in embryo selection means computers are increasingly influencing decisions about which children are born. He argued that patients with moral objections to AI involvement should be fully informed and ideally have the option to opt out. A review led by Koplin, published earlier this year in Human Reproduction, recommended greater oversight but concluded that ethical concerns alone are not sufficient to oppose AI in embryo assessment.
Michael Grynberg, a French obstetrician-gynaecologist specialising in IVF, stressed the need for more precise markers beyond traditional morphological assessments of eggs and sperm. “There is a lot of talk about AI,” he said. “We need better, more relevant markers to improve IVF outcomes.”
As AI continues to develop in reproductive medicine, it promises to make IVF more efficient, personalised, and potentially less emotionally taxing for parents. However, its use also demands careful ethical consideration and transparency, ensuring patients retain agency over these profoundly personal decisions.
