AI and IVF: Deep Learning in Embryo Selection

A study tested an AI deep learning system for embryo selection in IVF, comparing its choices with human experts. The AI matched the experts 65.8% of the time, achieving nearly the same pregnancy success rate. Despite its efficiency, proving the system's 'non-inferiority' requires larger trials.

An advanced AI deep learning system has been tested in the intricate procedure of embryo selection during in vitro fertilisation (IVF). Researchers aimed to compare the AI's performance with that of human embryologists. Published in Nature Medicine, the study presents compelling insights.

The system’s efficacy was evaluated in a randomized trial involving 1,066 IVF patients across 14 clinics in Australia and Europe. Both human experts and the AI selected embryos for implantation, followed by a randomized choice on which selection to use. The outcome showed a clinical pregnancy rate of 46.5% for AI-selected embryos and 48.2% for human-selected embryos, indicating minimal difference between the two.

While the AI demonstrated comparable accuracy, its embryo selection process was ten times quicker. However, proving 'non-inferiority' statistically would require a larger sample size. Notably, ethical concerns like the potential alteration of the sex ratio were addressed, showing no significant differences. This study suggests AI could enhance efficiency in embryology labs.