Artificial intelligence-driven oocyte scores are associated with early morphokinetics and embryo outcomes after...

How are the scores obtained using the deep learning-based oocyte assessment system, MAGENTA, associated with outcomes of intracytoplasmic sperm injection (ICSI) and biological events during the preimplantation period?

Retrospective study of 2785 images of mature oocytes from 1340 retrieval cycles. Images were obtained immediately after ICSI from time-lapse systems and scored by MAGENTA. Key phenomena, including pronuclear and cytoplasmic dynamics, cleavage patterns, blastomere compaction and embryo quality were manually annotated. The relationships between the MAGENTA score and embryonic and pregnancy outcomes, and morphokinetics were analysed using generalized linear models with generalized estimating equations to account for clustering.

Oocytes in the lower MAGENTA scores showed higher degeneration rates (P < 0.001) and lower normal fertilization rates (P < 0.001) per ICSI. Higher MAGENTA scores were associated with increased likelihood of development to the two- and eight-cell stages, compaction, blastulation and blastocyst expansion (all P < 0.001). Lower MAGENTA scores were associated with the following: delayed fertilization events, including second polar body extrusion, cytoplasmic wave initiation, halo redistribution and disappearance, and pronuclear appearance and breakdown (all P < 0.001); delayed first cleavage and higher incidences of abnormal cleavage, aberrant blastomere movement, and multinucleation (all P ≤ 0.001); and early compaction (P = 0.040), blastomere exclusion (P < 0.001) and poor inner cell mass and trophectoderm morphology (all P < 0.001). The sum of MAGENTA scores for all oocytes per retrieval cycle was positively associated with the cumulative live birth rate (P < 0.001).

The MAGENTA score is associated with key fertilization dynamics, cleavage abnormalities, compaction behaviour and blastocyst development, providing a quantitative indicator of oocyte developmental competence.