Some differences in early embryo development may be too subtle for human observation alone. A recent study published in Open Biology, “Using deep learning to predict the sex of human embryos,” explored this question using 515 human embryo time-lapse videos. The researchers combined manual annotation with deep learning analysis to investigate whether embryo development before implantation may contain patterns associated with birth sex.

The manual assessment did not provide a clear or reliable way to distinguish these patterns. However, the deep learning model showed statistically significant predictive performance, with reported accuracy around 61%. The authors also noted that developmental dynamics around and after the 8-cell stage may be particularly informative.

The study should not be read as a practical tool for embryo sex selection. The accuracy is limited, and the ethical implications of this topic require careful handling. Still, the paper is interesting because it points to a broader question in reproductive medicine: can AI detect developmental signals that are difficult for humans to recognize visually?

For IVF professionals, the most relevant takeaway may not be the specific prediction target itself, but the possibility that time-lapse data and AI could reveal subtle biological patterns that traditional observation does not easily capture. As AI continues to enter embryology, studies like this help expand the discussion from “Can AI score embryos?” to “What kinds of developmental information might AI be able to detect?”