O-106 Bringing Transparency to Oocyte Assessment: the importance of including confounders when building Artificial Intelligence (AI) based support tools to quantify oocyte viability

Abstract Study question Which confounders (sperm quality, oocyte dysmorphism, culture time, images pre or post-ICSI, age) affect the ability of AI to predict blastulation based on images? Summary answer Sperm post-ICSI image should be controlled for when building algorithms images. What is known already Previous studies reporting use have: (i) not accounted affecting (i.e. sperm time), and (ii) used images; without assessing whether ICSI procedure affects as assessed by AI. Therefore, there a risk mislabeling viable oocytes non-viable due external factors, which could cause uncontrolled bias failure generalize in clinical practice. The objective was assess how these efficacy prediction from an AI-based assessment tool: CHLOE-OQ(Fairtility). design, size, duration Cohort study. Images 1281 (February June 2022) were taken post using Embryoscope, embryos cultured until day 7. Oocyte donor source age, dysmorphias quality documented. CHLOE-OQ algorithm trained, validated tested diverse data set, accounting datasets, oocytes, patient age. Participants/materials, setting, methods primary endpoint blastulation. classified into 4 groups: (A)All (n = 1281), (B)donor only 51), (C)donor normospermic samples men diagnosed with male factor infertility 557), (D)abnormal other cycles 747). Eggs (own/donor), dysmorphisms: enlarged perivitelline space, abnormal Zona pellucida, cytoplasmic abnormalities, dark, oocytes. Main results role chance Post-ICSI had higher mean score than (0.28±0.1 vs 0.33±0.1, p < 0.001, paired t-test). Discrepancies particularly identified that degenerated following ICSI, scored 0 despite having scores pre-ICSI. Using (AUC=0.66, 95% confidence interval, CI: 0.63-0.69, n 1281) improved compared pre-ICSI (AUC=0.57: 0.53-0.60, 1281, 0.001), suggesting affected morphology oocyte, responds AI, contributes Efficacy (AUC) sperm: (A-OVERALL 0.658 [CI(95%): 0.626-0.687]; B-Donor 0.586 0.449-0.728]; C-normospermic 0.645 0.600-0.688], D 0.678 0.639-0.715]). features associated low were: dysmorphic abnormalities dark Whilst spherical normal zona space characterized being more likely form blastocyst. Limitations, reasons caution This single-clinic study retrospective. A multi-center underway. External factors must avoid good non-viable. There also need understand ensure biological transparency decision making. Wider implications findings Taking account gamete necessary strategy are generalized incorporated practice, whilst reducing promoting considering leads mislabeling, inaccurate predictions. Trial registration number applicable