L26/O-132 Artificial intelligence-based prediction of developmental competence in rescued metaphase II oocytes derived from germinal vesicle and metaphase I stages

Abstract Study question Can Magenta scores predict developmental competence and clinical outcomes of germinal vesicle (GV)- and metaphase-I (MI) stage-derived oocytes after rescue in vitro maturation (r-IVM)? Summary answer Magenta predicts stage-specific developmental competence of GV- and MI-derived metaphase-II (MII) oocytes after r-IVM, revealing intrinsically reduced developmental potential compared with in vivo-matured MII oocytes. What is known already Immature oocytes are frequently, albeit unintentionally, obtained following controlled-ovarian stimulation in assisted reproductive technology (ART). r-IVM has been used to increase the number of mature oocytes available, which can be particularly valuable in minimal-stimulation cycles. However, oocytes matured in vitro exhibit lower developmental competence than those matured in vivo. Moreover, morphological assessment of oocyte quality is subjective and unreliable, making it difficult to decide whether immature oocytes should be rescued. Artificial intelligence (AI) has recently been introduced in ART. Magenta, a deep-learning-based tool, predicts the blastocyst development potential of mature oocytes; however, its applicability to r-IVM MII oocytes requires validation. Study design, size, duration This retrospective study, conducted between March 2022 and June 2023, analysed oocytes retrieved from 765 minimal-stimulation cycles. A total of 2,480 oocytes (GV, n = 575; MI, n = 405; MII, n = 1,500) were evaluated using Magenta to provide scores (0-10) for in vivo-matured, in vitro-matured, and immature-stage oocytes. The associations between mean Magenta scores and maturation, fertilisation, blastocyst cryopreservation, and live birth rates were statistically assessed using multivariable logistic regression. Participants/materials, setting, methods Oocytes were denuded immediately following retrieval and cultured in a single-step embryo culture medium in time-lapse incubators (EmbryoScope+ or Flex). Immature oocytes (GV, MI) were cultured overnight for rIVM. Images were captured every 10 min, exporting the first and last frames to generate Magenta scores prior to intracytoplasmic sperm injection (ICSI) of MII oocytes. Following ICSI, all oocytes were cultured for 4–7 days until blastocyst cryopreservation. Single vitrified–warmed blastocyst transfers were performed. Main results and the role of chance Following r-IVM, maturation rates of GV- and MI-stage oocytes were 45.2% and 78.0%. Fertilisation rates were 81.2%, 75.6%, and 84.9% for GV-derived-, MI-derived-, and in vivo-matured MII oocytes (P = 0.0003). Blastocyst cryopreservation rates were lower in GV- and MI-derived MII oocytes (17.3% and 19.9%) than in in vivo-matured MII oocytes (45.1%; both P < 0.0001). Higher GV-stage Magenta scores (1.5±0.0) were positively associated with maturation and fertilisation (P = 0.0320–0.0390). Following maturation to MII, Magenta scores increased (2.1±0.1) and were predictive of blastocyst cryopreservation, with higher blastocyst development at scores ≥1.6 (24.8% vs < 1.6, 8.9%); no blastocysts were observed at scores