Robust and generalizable embryo selection based on artificial intelligence and time-lapse image sequences

Abstract

Assessing and selecting the most viable embryos for transfer is an essential part of in vitro fertilization (IVF). In recent years, several approaches have been made to improve and automate the procedure using artificial intelligence (AI) and deep learning. Based on images of embryos with known implantation data (KID), AI models have been trained to automatically score embryos related to their chance of achieving a successful implantation. However, as of now, only limited research has been conducted to evaluate how embryo selection models generalize to new clinics and how they perform in subgroup analyses across various conditions. In this paper, we investigate how a deep learning-based embryo selection model using only time-lapse image sequences performs across different patient ages and clinical conditions, and how it correlates with traditional morphokinetic parameters. The model was trained and evaluated based on a large dataset from 18 IVF centers consisting of 115,832 embryos, of which 14,644 embryos were transferred KID embryos. In an independent test set, the AI model sorted KID embryos with an area under the curve (AUC) of a receiver operating characteristic curve of 0.67 and all embryos with an AUC of 0.95. A clinic hold-out test showed that the model generalized to new clinics with an AUC range of 0.60–0.75 for KID embryos. Across different subgroups of age, insemination method, incubation time, and transfer protocol, the AUC ranged between 0.63 and 0.69. Furthermore, model predictions correlated positively with blastocyst grading and negatively with direct cleavages. The fully automated iDAScore v1.0 model was shown to perform at least as good as a state-of-the-art manual embryo selection model. Moreover, full automatization of embryo scoring implies fewer manual evaluations and eliminates biases due to inter- and intraobserver variation.

Correlation between an annotation-free embryo scoring system based on deep learning and live birth/neonatal outcomes after single vitrified-warmed blastocyst transfer: a single-centre, large-cohort retrospective study

“…For example, embryo developmental speed, morphokinetic parameters, and blastocyst grade [ 21 , 40 , 46 , 47 ] have been reported to correlate with both sex ratio and implantation likelihood. In a study on iDAScore, it was shown that scores correlate with both time to blastocyst and blastocyst grades [ 30 ]. This could possibly be one of the causes for the observed non-significant trend in the male sex ratio.…”

“…This is different from the current study where the test data set only includes transferred embryos and specifically discarded embryos are not included in the AUC calculation. In the analysis of iDAScore v1.0 by Berntsen et al [ 30 ], the AUC for FHB + vs (FHB − and discarded) was 0.95 and for FHB + vs FHB − the AUC was 0.67. The differences between these two AUC values clearly show that it is much easier to discriminate between FHB + and discarded embryos than to discriminate among transferred embryos alone.…”

“…The iDAScore v1.0 model [ 30 ] was developed using a 3D convolutional neural network and was trained on time-lapse sequence images [ 19 ]. The only input to the model was the images of a time-lapse sequence, and the output was a numerical score of 1–9.9, correlating with the likelihood of obtaining a positive foetal heartbeat.…”

An artificial intelligence model correlated with morphological and genetic features of blastocyst quality improves ranking of viable embryos

“…These values, while allowing a general comparison with similar AIs described in the literature (e.g. Alife Health AI ROC-AUC 0.62-0.64, Fairtility AI ROC-AUC 0.68-0.70, Vitrolife AI ROC-AUC 0.67see Loewke et al, 2022, Erlich et al, 2022, and Berntsen et al, 2022, do not give an indication of the ability of the AI model to effectively rank embryos for selection within a patient cohort during a single IVF cycle.…”

“…In recent years there has been a growing interest in the use of artificial intelligence (AI) to support embryo quality assessment, with numerous algorithms being developed for analysis of static images or time-lapse videos of embryos to aid in the selection of embryos for transfer (Berntsen et al, 2022;Chavez-Badiola et al, 2020;Erlich et al, 2022;Khosravi et al, 2019;Loewke et al, 2022;Silver et al, 2020;Tran et al, 2019;VerMilyea et al, 2020). However, due to the infancy of AI in the embryology field, few of these studies have evaluated real-world clinical use.…”