Abstract Study question Can AI reduce the rate of twin and triplet monozygotic pregnancies following elective single embryo transfer (eSET)? Summary answer AI can automatically annotate morphokinetic developments and other biomarkers. Embryos leading multiple monozygotic pregnancies have slower cell divisions and larger ICM than monoamniotic embryos. What is known already Monozygotic twin (MZT) and monozygotic triplet (MTP) pregnancies are a rare phenomenon in spontaneous pregnancies, but the incidence increases significantly when pregnancies are achieved by assisted reproductive technology (ART); 0.4% vs. 1.56% MZTs, and 0.004% vs. 0.048% MTPs. It is unclear what mechanisms cause an embryo to split into two or three, although several have been proposed such as culture to blastocyst, decompacting ICMs, frozen-warmed embryo transfers, assisted hatching and even ICSI. In order to study this split phenomenon, time-lapse imaging has been used to discover any signs of embryo division. Study design, size, duration This is a retrospective assessment of a total of 8 embryos from 2018 to 2022 that led to single pregnancy (n = 4), twin pregnancy (n = 3) and triple pregnancy (n = 1) following a fresh single embryo transfer. All embryos were inseminated using the ICSI technique and assisted hatching was performed before fresh transfer. Participants/materials, setting, methods Using CHLOE (Fairtility, Tel Aviv) we automatically assessed the morphokinetics, blastocyst biomarkers and scores. Singletons and multiple pregnancies were compared in terms of morphokinetics (t-test) and surface area of ICM, ICM diameter, ICM area/embryo area ratio, ICM shape and CHLOE embryo quality score were compared using the ANOVA test. Main results and the role of chance Embryos that led to multiple pregnancies had slower embryo development than embryos that led to singletons (single vs multiple: t2: 22.1+-2 vs 25.2+-2, p = 0.04; t3: 32.56+-2.65 vs 36.98 +-1.48, p = 0.01; t5: 44.09+-4.84 vs 50.37+−0.69, p = 0.02; t6: 47.26+-3.87 vs 53.59+-2.40, p = 0.01; t7: 48.76+-4.04 vs 55.15+-3, p = 0.02; t8: 50.18+-4.27 vs 57.55+-1.5, p = 0.008; t9: 63.27+-3.76 vs 73.04+-5.33, p = 0.03). Embryos leading to triplets were slower than twins which were, in turn, slower than singletons (single vs twins vs triplets: t4: 33.86+-3.41 vs 37.69+-1.61 vs 48.95, p = 0.007; t8: 50.18+-4.27 vs 58.07+-1.29 vs 55.96, p = 0.04; t9: 63.27+-3.76 vs 71.23+-4.81 vs 78.44, p = 0.01). Embryos that led to multiple pregnancies had a larger ICM to embryo surface area ratio (single vs multiple: 0.14+−0.08 vs 0.27+−0.06, p = 0.04) and smaller embryo diameter (single vs multiple: 177.2+-16.5 vs 138.15+-8.25, p = 0.003). We didn’t find statistically significant differences between the groups in CHLOE EQ score (single vs multiple: 0.97 vs 0.76, p = 0.36), Blast score (single vs multiple: 0.87+−0.06 vs 0.78+−0.13, p = 0.26), CHLOE Rank, Trophectoderm quality, ICM Area (single vs multiple: 3629+-1361 vs 4151 +- 569, p = 0.48) and ICM shape (single vs multiple: 1.64+−0.45 vs 1.27+−0.12, p = 0.15). Limitations, reasons for caution The main limitation of this study is the number of cases included in the study, as we studied one triplet, 3 twins and 4 single pregnancies. Wider implications of the findings This might be the first time that AI has been used to analyse the behaviour of embryos resulting in multiple pregnancies. The transfer of a slowly dividing embryo and/or with a large ICM could result in a multiple pregnancy. Trial registration number not applicable
Read full abstract