Deep learning analysis reveals distinct expansion patterns between euploid and aneuploid embryos during blastulation

Abstract

Study question Can deep learning analysis of embryo expansion patterns from time-lapse imaging identify differences between euploid and aneuploid embryos?

Summary answer Aneuploid embryos showed slower expansion rates and more frequent deflation episodes compared to euploid embryos during blastulation

What is known already Embryo expansion patterns during blastulation are clinically significant indicators of developmental potential. Current assessment methods rely on subjective categorical classifications, limiting their predictive value. Time-lapse imaging provides continuous embryo development tracking, yet manual annotation of expansion kinetics is impractical. Previous studies have shown associations between expansion characteristics and clinical outcomes, but objective quantification methods are lacking. The relationship between chromosomal status and expansion patterns remains poorly understood.

Study design, size, duration Retrospective analysis of 418 time-lapse videos from embryos that underwent preimplantation genetic testing for aneuploidy (PGT-A), including 140 euploid and 278 aneuploid embryos. Deep learning segmentation models were used to analyze expansion patterns. Videos were processed to extract frames at 5 per second, beginning at the 15-second mark corresponding to 75 hours post-insemination (h.p.i).