JBRA Assist. Reprod. 2025;29(Suppl 1):20-20
POSTER PRESENTATION
doi: 10.5935/1518-0557.20250075
1FERTILIA - Reproductive Medicine
2FAIRTILITY
Objective: To assess if an artificial intelligence (AI) algorithm has better prediction than an experienced embryologist to estimate the blastocyst stage development. To assess if the AI scores obtained during embryo culture correlates with embryo development, male factor, blastocyst morphology grading and ploidy.
Methods: Retrospective cohort study (197 embryos derived from 82 egg donation cycles). Embryos were assessed at 30,72,120,144 and 168 hpi (hours post-insemination) during time lapse incubation. The blastocysts stage (BL) prediction was assessed using an AI algorithm at 30hpi and 72hpi and compared with experienced embryologists’ prediction at 72hpi. The embryo quality score (EQ) correlation with embryo development, blastocysts morphology and ploidy were also assessed. Ploidy analysis was performed in 56 blastocysts. The male factor (MF) impact over the BL and EQ scores was evaluated. Specificity (S), sensibility (SE) and area under the curve (AUC) examination were performed.
Results: AThe embryologist blastocyst prediction was AUC 0.72 (S:78.7%-SE:66 %) compared with AI 0.68 (S:71.1%-SE:58.5%) and 0.81 (S:88.7-SE:74%) at 30hpi and 72hpi, respectively. Blastocysts with high morphology grading had higher EQ scores at 72hpi compared with good and poor quality (7.8 vs 5.9 and 5.1, respectively, p<0.05.). The EQ score at 72hpi to predict ploidy was AUC 0.87 with SE:100% and S:61.4%. Cycles with MF does not have different BL or EQ scores at 30hpi or 72hpi.
Conclusion: The blastocyst stage prediction is 10% better considering AI in relation with senior embryologists at 72hpi. The AI EQ scores had a good correlation with day 5, high quality morphology and ploidy.MF is not associated with different AI EQ scores.