The retrospective cohort study involved 979 TL-PGT cycles conducted from 2018 to 2021 at the Reproductive Medicine Center, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, Hubei, China. A total of 3448 blastocysts were biopsied. These samples were lysed, and the DNA was fragmented and amplified. Among them, 3405 samples were amplified successful, and 42 samples were failed to amplify. In this work, the iDAScore model was used to retrospectively examine 3405 blastocysts that were cultivated in an EmbryoScope Plus (Vitrolife A/S, Denmark) incubator. In Fig. 1, the research design is displayed. On an informed consent form, each patient signed. Every patient signed an informed consent form. The study complied with the Declaration of Helsinki for Human Subjects in Medical Research and the Board of Institutional Review (No. 2019s097) approval was given by the Ethical Committee of Reproductive Medicine Center, Tongji Hospital, Tongji Medicine College, Huazhong University of Science and Technology.

Schematic presentation of the study design. TL, time-lapse; PGT, preimplantation genetic testing

Controlled ovarian stimulation (COS) was performed in compliance with our previous studies [16]. Patients were regularly monitored with transvaginal ultrasonography during COS. When the leading follicle(s) measured more than 18 mm, human chorionic gonadotropin (HCG) was administered. The oocytes were retrieved 36 h after the HCG injection, guided by ultrasonography. The collection of cumulus-oocyte complexes (COCs) was performed as described in our previous studies [17].

The density gradient centrifugation method was used to optimize the semen samples [18]. The sperm concentration, motility, and morphology were evaluated using the fifth edition of the World Health Organization recommendations. During intracytoplasmic sperm injection (ICSI), the COCs were denuded two hours after retrieval, and sperm was injected four hours later. The resultant zygotes were subsequently cultured utilizing a time-lapse incubation system at G1 Plus (Vitrolife, Sweden). Each embryo was photographed every ten minutes in TL incubator. After insemination, pronuclei were inspected 16–18 h later. The culture medium was switched to G2 Plus (Vitrolife, Sweden) on the third day. The blastocysts that met the Gardner criterion (better than 3BC) were biopsied and then cryopreserved for later use on the fifth and sixth day. Rarely, the embryo was cultivated for vitrification up until the seventh day. Before performing a biopsy, a tiny hole in the zona pellucida is created using a laser. This allowed for the mechanical dissection of three to six trophectoderm cells.

The PGT cycles were conducted with NGS analysis [19]. To summarize, the samples were amplified using a single-cell whole-genome amplification (WGA) based on multiple annealing and looping-based amplification cycles (MALBACs), in accordance with commercial kit protocol (Yikon Genomics). DNA was fragmented, amplified, labelled, and purified in a sequential manner. Utilizing Life Technologies’ Ion Proton technology, the final library was sequenced at a depth of around 0.04× genomes. In order to detect variants, this sequencing speed generates repeatable copy number variations (CNVs) at ∼ 4 MB resolution. A threshold of more than 70% was established for the detection of aneuploidy. When it comes to chromosomes, the threshold for mosaic detection differs. The lower limit was 30% for chromosomes 13, 16, 18, and 21, 50% for chromosome 19, and 40% for all other chromosomes. A number that is below the lower bound denotes euploidy.

A deep learning neural network that is trained on TL videos to predict fetal heartbeat is the iDAScore embryo scoring model. Time-lapse videos are fed into the iDAScore model, which generates an embryo score between 1.0 and 9.9. The data from the blastocysts that were included in this study were assessed retrospectively using the iDAScore model.

Continuous variables with normal distributions were expressed as mean ± SD. Categorical variables were expressed as number and percentage (%). For data with a normal distribution, one-way analysis of variance (ANOVA) was applied for multiple comparisons. The chi-square test was used to compare categorical variables between groups. Multivariable logistic regression was applied to evaluate the association between the iDAScores and ploidy, and the odds ratios (ORs) were calculated. All statistical tests were two-sided and p values less than 0.05 were considered statistically significant. All analyses were conducted in SPSS Statistics (version 23.0, IBM, Armonk, NY, USA).