P-280 Potential for improvement and current limitations of Artificial Intelligence (AI) for embryo selection: analysis of external validation data

Study question

What are the prospects of improvement and the limitations of an AI system for embryo selection?

Summary answer

The predictive performance of AI can be enhanced by including additional factors, on top of embryo images, and by assessing images with centered blastocysts.

What is known already

We previously reported the external validation of IVFvision.ai, an AI algorithm that differentiates between Day-5 blastocysts with a positive or negative implantation outcome. IVFvision.ai had higher AUC and overall accuracy in predicting implantation compared to KIDScoreD5 and senior embryologists. Here we report a secondary analysis of external validation data, focusing on a) the improvement of the predictive ability of IVFvision.ai by incorporating data from additional sources, and b) the impact of the blastocyst image quality on the performance of IVFvision.ai.

Study design, size, duration

This is a secondary analysis of external validation data. External validation of IVFvision.ai was performed at a University IVF Clinic using 113 anonymised Embryoscope images of single D5 blastocyst transfers with known implantation outcome.

Participants/materials, setting, methods

The performance of IVFvision.ai and three senior Embryologists to correctly classify blastocysts according to implantation outcome were compared in images in which the whole blastocyst was visible (centred blastocysts, n = 62) vs images in which part of the blastocyst was not visible (off-centred blastocysts, n = 51). Logistic regression models were created: a) IVFvision alone, b) IVFvision+age, c) IVFvision+fertilisation_method, d) IVFvision+KIDScoreD5, e) IVFvision+age+Fertilisation_method+KIDScoreD5. The AUC of each model in predicting implantation was estimated using ROC curve analysis.

Main results and the role of chance

The AUC of IVFVision.ai (0.675 vs 0.432), Embryologist 1 (0.570 vs 0.390), Embryologist 2 (0.663 vs 0.448) and Embryologist 3 (0.628 vs 0.485) were higher for images with centered blastocysts compared to non-centered blastocysts, respectively. There was a progressive increase of AUC with the addition of more factors in the predictive models. a) IVFvision alone: AUC=0.675, b) IVFvision+age: AUC=0.675 c) IVFvision+KIDScoreD5: AUC=0.721 d) IVFvision+fertilisation_method=0.740, e) IVFvision+age+Fertilisation_method+KIDScoreD5=0.768.

Limitations, reasons for caution

The retrospective nature of the study and the small sample of the study raise the need for further prospective studies with a larger number of embryos.

Wider implications of the findings

The highest performance of IVFvision.ai is achieved in images with centred blastocysts, suggesting that implantation cannot be predicted accurately in images with non-centred blastocysts. In addition, we provide provide proof of concept that training AI systems using data from different sources, in addition to embryo images, may increase overall accuracy.

Trial registration number

not applicable

P–755 Perinatal outcomes following Day–4 embryo transfer compared to Day–2, Day–3 and Day–5 embryo transfer: an analysis of 56,346 singleton live births

Study question

Are perinatal outcomes of singleton live births following Day–4 embryo transfer (ET) different to Day–2, Day–3 and Day–5 ET?

Summary answer

Perinatal outcomes of singleton live births following Day–4 ET are similar with those following Day–2, Day–3 and Day–5 ET.

What is known already

The morula represents a critical stage in preimplantation embryo development, but the usage of morula transfer on Day–4 has received little attention. Recent work from our group suggested that live birth rates following Day–4 ET appear higher than cleavage-stage ET, but lower than blastocyst ET. Therefore, Day–4 appears an alternative day to perform ET, offering the benefits of extended culture for embryo selection, but with shorter in-vitro culture exposure, as well as improving flexibility and planning in the IVF Clinic. However, there are extremely limited data available on the perinatal outcomes after Day–4 ET compared to cleavage-stage and blastocyst ET.

Study design, size, duration

Retrospective cohort study using data from the anonymised dataset of the Human Fertilisation and Embryology Authority (HFEA) in the UK between 2000 and 2016. Data from singleton live births of women undergoing their first IVF/ICSI cycle were analysed to compare perinatal outcomes after fresh Day–2,3,4,5 embryo transfers.

Participants/materials, setting, methods

Births resulting from the first, fresh, autologous, stimulated, non-PGT cycles, with full data, were included. After exclusions, a total 56,346 singleton live births were included in the analysis (17,613 from Day–2 ET, 15,533 from Day–3 ET, 508 from Day–4, 22,692 from Day–5 ET).

Binary/multinomial logistic regression analysis was performed to adjust for important cofounders. Adjusted odds ratios (aORs) and 95% confidence intervals (95%CI) were calculated. The level of significance was set at < 0.05.

Main results and the role of chance

The probabilities of birth at full-term (FT) and normal birthweight (NBW) after Day–4 transfer (FT 90.4%; NBW 84.6%) were similar to Day–2 (FT 89.7%, aOR 0.994, [0.734–1.344]; NBW 81.9%, aOR 0.881, [0.708–1.096]), Day–3 (FT 90.2%, aOR 1.026, [0.760–1.386]; NBW 82.4%, aOR 0.894, [0.719–1.111]) and Day–5 transfer (FT 90.4%, aOR 1.001, [0.743–1.350]; NBW 83.7%, aOR 0.920, [0.741–1.142]).

The probabilities of preterm birth (PTB) and very preterm birth (VPTB) after Day–4 transfer (PTB 9.3%; VPTB 0.4%) were similar to Day–2 (PTB 9.5%; aOR=0.952; VPTB 0.8%; aOR=2.172), Day–3 (PTB 9.0%, aOR=0.920; VPTB 0.9%, aOR=2.174), and Day–5 transfer (PTB 8.8%; aOR=0.955; VPTB 0.8%, aOR=1.956).

The probabilities of very-low birthweight (VLBW), low birthweight (LBW), high birthweight (HBW) and very-high birthweight (VHBW) after Day–4 transfer (VLBW 0.9%, LBW 7.9%, HBW 6.3%, VHBW 0.3%) were similar to Day–2 (VLBW 1.8%, aOR=1.827; LBW 8.0%, aOR=1.015; HBW 8.1%, aOR=1.174; VHBW 0.2%, aOR=0.590), Day–3 (VLBW 1.8%, aOR=1.788; LBW 7.4%, aOR=0.927; HBW 8.3%, aOR=1.256; VHBW 0.2%, aOR=0.503) and Day–5 transfer (VLBW 1.6%, aOR=1.782; LBW 6.9%, aOR=0.894; HBW 7.5%, aOR=1.215; VHBW 0.2%, aOR=0.796).

The probability of having a female baby after Day–4 transfer (51.6%) was similar to Day–2 (49.2%, aOR 0.940), Day–3 (49.3%, aOR 0.931) and Day–5 transfer (48.3%, aOR 0.869).

Limitations, reasons for caution

The study is limited by its retrospective nature, the inability to adjust for additional confounders and the small number of singleton births after Day–4 ET. It is not known how Day–4 ET was decided. The incidence of congenital abnormalities was not analysed due to incomplete registration in the dataset.

Wider implications of the findings: Perinatal outcomes of singleton live births following Day–4 ET are similar with those following Day–2, Day–3 and Day–5 ET, suggesting that morula transfer is equally safe as cleavage-stage and blastocyst transfer. Data on a larger number of live births from well-designed RCTs are required to confirm these findings.

Trial registration number

Not applicable