AI performs a bit worse than humans in selecting viable IVF embryos

Embargoed until: Publicly released: 2024-08-09 19:00

Australia; NSW; VIC

CC-0. https://picryl.com/media/embryo-ivf-icsi-5aacd4

Australian-led research compared the ability of a deep learning artificial intelligence (AI) to pick viable embryos for in vitro fertilisation (IVF) with the performance of a standard assessment by a professional, and found there wasn't much difference between the two, although the human outperformed the AI slightly. Across 14 in vitro fertilisation clinics in Australia and Europe, 1,066 women with at least two early-stage blastocysts on day 5 used either a standard morphological assessment or a deep learning algorithm to select their most viable embryos. Of the group using AI, the pregnancy rate was 46.5%, while with the standard assessment, the pregnancy rate was 48.2%.

Attachments

Note: Not all attachments are visible to the general public. Research URLs will go live after the embargo ends.

Research Springer Nature, Web page The URL will go live after the embargo ends

Journal/
conference: Nature Medicine

Research:Paper

Organisation/s: IVFAustralia, The University of New South Wales, Melbourne IVF, The University of Melbourne, Virtus Health

Funder: The project was funded by a grant from Vitrolife. As sponsor, employees of Vitrolife Group participated in the study design and writing of the report. Data collection and analysis were performed by independent entities.

Media Contact/s

Contact details are only visible to registered journalists.