O-239 A model based on artificial intelligence for the non-invasive prediction of embryo aneuploidy

Abstract Study question Can the automated artificial intelligence (AI) model iDAScore predict embryo aneuploidy? Summary answer The AUC for aneuploidy prediction of was 0.640, with optimal cut-off score 7.75, sensitivity 61.0% and specificity 60.9%. What is known already Embryo morphokinetics image analysis using AI have been proposed as alternative approaches non-invasive prediction. iDAScore, well morphokinetic selection KIDScoreD5 shown to correlation ploidy status. aim study performance morphology grading (Gardner criteria). design, size, duration Retrospective including 382 blastocysts from 92 patients undergoing PGT-A in a single centre during January-December 2022. For statistical analysis, continuous categorical data were compared ANOVA Fisher’s exact test/Chi-square, respectively. ROC used estimate discrimination variables aneuploidy. Stepwise logistic regression identify confounding construct Participants/materials, setting, methods Embryos cultured Embryoscope, underwent laser assisted hatching on Day3, biopsied Day5-Day6 vitrified. Whole-genome amplification next-generation sequencing PGT-A. values calculated v1.2.0. v.3 after annotation required parameters Embryoviewer. Gardner criteria, inner cell mass (ICM) trophectoderm (TE) classified into grades A C. expansion all > =grade 4. Main results role chance Out embryos analysed, 127 (33.2%) euploid 255 (66.8%) aneuploid. Euploid associated significantly higher (7.69, 95% CI 7.44-7.95 vs 6.86, 6.66-7.09, p < 0.001), (6.41, 6.07-6.75 5.74, 5.51-5.97, = 0.001) lower maternal age (37.10 years, 36.39-37.82 39.83 39.36-40.30, aneuploid embryos. Significantly more grade ICM (A:39.4%, B:26.0%, C:20.0%, 0.013) TE (A: 45.9%, B:29.6%, C:23.6%, 0.002) B C grades, Blastocysts Day 5 6 had similar proportions euploidy (34.9% 27.6%, 0.125). AUCs were: iDAScore: 0.640 (95% 0.582-0.699), KIDScoreD5: 0.604 0.542-0.665), grade: 0.581 0.521-0.640), 0.596 (0.536-0.657), age: 0.694 (0.640-0.748). (Youdex index) 7.75 (sensitivity 61.0%, 60.9%). Following stepwise multivariate regression, that entered grade. iDAScore+age+TE increased ability (AUC 0.747, 0.696-0.796). KIDScoreD5, day biopsy failed enter model. Limitations, reasons caution This retrospective therefore presence bias cannot be excluded. KIDScore are models originally developed implantation not diagnosis Wider implications findings appears associate Combination can decision-support tool prioritising transfer, cryopreservation or biopsy. needed confirm present findings. Trial registration number applicable