Fertilization signatures as biomarkers of embryo quality

A deep learning model for predicting blastocyst formation from cleavage-stage human embryos using time-lapse images

Efficient prediction of blastocyst formation from early-stage human embryos is imperative for improving the success rates of assisted reproductive technology (ART). Clinics transfer embryos at the blastocyst stage on Day-5 but Day-3 embryo transfer offers the advantage of a shorter culture duration, which reduces exposure to laboratory conditions, potentially enhancing embryonic development within a more conducive uterine environment and improving the likelihood of successful pregnancies. In this paper, we present a novel ResNet-GRU deep-learning model to predict blastocyst formation at 72 HPI. The model considers the time-lapse images from the incubator from Day 0 to Day 3. The model predicts blastocyst formation with a validation accuracy of 93% from the cleavage stage. The sensitivity and specificity are 0.97 and 0.77 respectively. The deep learning model presented in this paper will assist the embryologist in identifying the best embryo to transfer at Day 3, leading to improved patient outcomes and pregnancy rates in ART.

Zygotic spindle orientation defines cleavage pattern and nuclear status of human embryos

The first embryonic division represents a starting point for the development of a new individual. In many species, tight control over the first embryonic division ensures its accuracy. However, the first division in humans is often erroneous and can impair embryo development. To delineate the spatiotemporal organization of the first mitotic division typical for normal human embryo development, we systematically analyzed a unique timelapse dataset of 300 IVF embryos that developed into healthy newborns. The zygotic division pattern of these best-quality embryos was compared to their siblings that failed to implant or arrested during cleavage stage. We show that division at the right angle to the juxtaposed pronuclei is preferential and supports faithful zygotic division. Alternative configurations of the first mitosis are associated with reduced clustering of nucleoli and multinucleation at the 2-cell stage, which are more common in women of advanced age. Collectively, these data imply that orientation of the first division predisposes human embryos to genetic (in)stability and may contribute to aneuploidy and age-related infertility.

Two types of cleavage, from zygote to three cells, result in different clinical outcomes and should be treated differently

Research Question

What is the utilization rate of embryos that exert inadequate zygote cleavage into three daughter cells?

Design

This study used a retrospective dataset from a single IVF Unit. A total of 3,060 embryos from 1,811 fresh IVF cycles were analyzed. The cleavage pattern, morphokinetics, and outcome were recorded. Only 2pn embryos, fertilized by ejaculated sperm, and cultured in a time-lapse system for at least 5 days were included. We generated three study groups according to the embryo's cleavage pattern: (I) Control, normal cleavage (n = 551); (II) fast cleavage, zygote to three cells within 5 h (n = 1,587); and (III) instant direct tripolar cleavage (IDC) from zygote to three cells (n = 922).

Results

The rate of usable fast cleavage blastocysts was 108/1,587 (6.81%) and usable control blastocysts was 180/551 (32.67%). The time of PN fading and from fading to first cleavage differed significantly between the three groups. Although the pregnancy rate of control and fast cleavage blastocysts were comparable (40.35% and 42.55%, respectively), the amount of instant direct cleavage embryos that reached blastocyst stage was neglectable (only four embryos out of 922 analyzed IDC embryos) and unsuitable for statistical comparison of pregnancy rates.

Conclusion

Our results indicate the need to culture instant direct cleavage embryos for 5 days, up to the blastocyst stage, and avoid transfer of embryos that are fated to arrest even when their morphological grade on day 3 is acceptable, whereas fast cleavage embryos could be transferred on day 3 when there is no alternative.

Time-lapse imaging of morula compaction for selecting high-quality blastocysts: a retrospective cohort study

PURPOSE

Before blastocyst development, embryos undergo morphological and metabolic changes crucial for their subsequent growth. This study aimed to investigate the relationship between morula compaction and blastocyst formation and the subsequent chromosomal status of the embryos.

METHODS

This retrospective cohort study evaluated embryo development (n = 371) using time-lapse imaging; 94 blastocysts underwent preimplantation genetic testing for aneuploidy (PGT-A).

RESULTS

The embryos were classified as fully (Group 1, n = 194) or partially (Group 2, n = 177) compacted. Group 1 had significantly higher proportions of good- and average-quality blastocysts than Group 2 (21.6% vs. 3.4%, p = 0.001; 47.9% vs. 26.6%, p = 0.001, respectively). The time from the morula stage to the beginning and completion of compaction and blastocyst formation was significantly shorter in Group 1 than in Group 2 (78.6 vs. 82.4 h, p = 0.001; 87.0 vs. 92.2 h, p = 0.001; 100.2 vs. 103.7 h, p = 0.017, respectively). Group 1 embryos had larger surface areas than Group 2 embryos at various time points following blastocyst formation. Group 1 blastocysts had significantly higher average expansion rates than Group 2 blastocysts (653.6 vs. 499.2 μm2/h, p = 0.001). PGT-A revealed a higher proportion of euploid embryos in Group 1 than in Group 2 (47.2% vs. 36.6%, p = 0.303).

CONCLUSION

Time-lapse microscopy uncovered a positive relationship between compaction and blastocyst quality and its association with embryo ploidy. Hence, compaction evaluation should be prioritized before blastocyst selection for transfer or cryopreservation.

Developmental perturbation in human embryos: Clinical and biological significance learned from time-lapse images

Background

Time-lapse technology (TLT) has gained widespread adoption worldwide. In addition to facilitating the undisturbed culture of embryos, TLT offers the unique capability of continuously monitoring embryos to detect spatiotemporal changes. Although these observed phenomena play a role in optimal embryo selection/deselection, the clinical advantages of introducing TLT remain unclear. However, manual annotation of embryo perturbation could facilitate a comprehensive assessment of developmental competence. This process requires a thorough understanding of embryo observation and the biological significance associated with developmental dogma and variation. This review elucidates the typical behavior and variation of each phenomenon, exploring their clinical significance and research perspectives.

Methods

The MEDLINE database was searched using PubMed for peer-reviewed English-language original articles concerning human embryo development.

Main findings

TLT allows the observation of consecutive changes in embryo morphology, serving as potential biomarkers for embryo assessment. In assisted reproductive technology laboratories, several phenomena have not revealed their mechanism, posing difficulties such as fertilization deficiency and morula arrest.

Conclusion

A profound understanding of the biological mechanisms and significance of each phenomenon is crucial. Further collaborative efforts between the clinical and molecular fields following translational studies are required to advance embryonic outcomes and assessment.

A hybrid artificial intelligence model leverages multi-centric clinical data to improve fetal heart rate pregnancy prediction across time-lapse systems

STUDY QUESTION

Can artificial intelligence (AI) algorithms developed to assist embryologists in evaluating embryo morphokinetics be enriched with multi-centric clinical data to better predict clinical pregnancy outcome?

SUMMARY ANSWER

Training algorithms on multi-centric clinical data significantly increased AUC compared to algorithms that only analyzed the time-lapse system (TLS) videos.

WHAT IS KNOWN ALREADY

Several AI-based algorithms have been developed to predict pregnancy, most of them based only on analysis of the time-lapse recording of embryo development. It remains unclear, however, whether considering numerous clinical features can improve the predictive performances of time-lapse based embryo evaluation.

STUDY DESIGN, SIZE, DURATION

A dataset of 9986 embryos (95.60% known clinical pregnancy outcome, 32.47% frozen transfers) from 5226 patients from 14 European fertility centers (in two countries) recorded with three different TLS was used to train and validate the algorithms. A total of 31 clinical factors were collected. A separate test set (447 videos) was used to compare performances between embryologists and the algorithm.

PARTICIPANTS/MATERIALS, SETTING, METHODS

Clinical pregnancy (defined as a pregnancy leading to a fetal heartbeat) outcome was first predicted using a 3D convolutional neural network that analyzed videos of the embryonic development up to 2 or 3 days of development (33% of the database) or up to 5 or 6 days of development (67% of the database). The output video score was then fed as input alongside clinical features to a gradient boosting algorithm that generated a second score corresponding to the hybrid model. AUC was computed across 7-fold of the validation dataset for both models. These predictions were compared to those of 13 senior embryologists made on the test dataset.

MAIN RESULTS AND THE ROLE OF CHANCE

The average AUC of the hybrid model across all 7-fold was significantly higher than that of the video model (0.727 versus 0.684, respectively, P = 0.015; Wilcoxon test). A SHapley Additive exPlanations (SHAP) analysis of the hybrid model showed that the six first most important features to predict pregnancy were morphokinetics of the embryo (video score), oocyte age, total gonadotrophin dose intake, number of embryos generated, number of oocytes retrieved, and endometrium thickness. The hybrid model was shown to be superior to embryologists with respect to different metrics, including the balanced accuracy (P ≤ 0.003; Wilcoxon test). The likelihood of pregnancy was linearly linked to the hybrid score, with increasing odds ratio (maximum P-value = 0.001), demonstrating the ranking capacity of the model. Training individual hybrid models did not improve predictive performance. A clinic hold-out experiment was conducted and resulted in AUCs ranging between 0.63 and 0.73. Performance of the hybrid model did not vary between TLS or between subgroups of embryos transferred at different days of embryonic development. The hybrid model did fare better for patients older than 35 years (P < 0.001; Mann-Whitney test), and for fresh transfers (P < 0.001; Mann-Whitney test).

LIMITATIONS, REASONS FOR CAUTION

Participant centers were located in two countries, thus limiting the generalization of our conclusion to wider subpopulations of patients. Not all clinical features were available for all embryos, thus limiting the performances of the hybrid model in some instances.

WIDER IMPLICATIONS OF THE FINDINGS

Our study suggests that considering clinical data improves pregnancy predictive performances and that there is no need to retrain algorithms at the clinic level unless they follow strikingly different practices. This study characterizes a versatile AI algorithm with similar performance on different time-lapse microscopes and on embryos transferred at different development stages. It can also help with patients of different ages and protocols used but with varying performances, presumably because the task of predicting fetal heartbeat becomes more or less hard depending on the clinical context. This AI model can be made widely available and can help embryologists in a wide range of clinical scenarios to standardize their practices.

STUDY FUNDING/COMPETING INTEREST(S)

Funding for the study was provided by ImVitro with grant funding received in part from BPIFrance (Bourse French Tech Emergence (DOS0106572/00), Paris Innovation Amorçage (DOS0132841/00), and Aide au Développement DeepTech (DOS0152872/00)). A.B.-C. is a co-owner of, and holds stocks in, ImVitro SAS. A.B.-C. and F.D.M. hold a patent for 'Devices and processes for machine learning prediction of in vitro fertilization' (EP20305914.2). A.D., N.D., M.M.F., and F.D.M. are or have been employees of ImVitro and have been granted stock options. X.P.-V. has been paid as a consultant to ImVitro and has been granted stocks options of ImVitro. L.C.-D. and C.G.-S. have undertaken paid consultancy for ImVitro SAS. The remaining authors have no conflicts to declare.

TRIAL REGISTRATION NUMBER

N/A.

Human 1PN and 3PN zygotes recapitulate all morphokinetic events of normal fertilization but reveal novel developmental errors

STUDY QUESTION

Does mono- (1PN) and tri-pronuclear (3PN) fertilization recapitulate the morphokinetic changes of normal bi-pronuclear (2PN) fertilization?

SUMMARY ANSWER

Abnormal fertilization retraces the overall choreography of normal fertilization but reveals novel morphokinetic phenomena and raises scientifically and clinically relevant questions.

WHAT IS KNOWN ALREADY

ART has allowed the extracorporeal observation of early human development. Time-lapse technology (TLT) has revealed the complexity of the morphokinetic changes underpinning fertilization and the importance of this process for the genetic and cellular integrity of the embryo. Abnormal fertilization has remained neglected, despite its relevance to the physiology and pathology of early human development.

STUDY DESIGN, SIZE, DURATION

This retrospective study involved TLT observation of normally (2PN, N = 2517) and abnormally (1PN, N = 41; 3PN, N = 27) fertilized oocytes generated in ICSI cycles performed between October 2019 and December 2020. Oocyte retrieval was carried out after clomiphene citrate-based minimal ovarian stimulation. Oocytes of patients with different diagnoses of infertility were included in the analysis, while cases involving cryopreserved gametes or surgically retrieved sperm were excluded.

PARTICIPANTS/MATERIALS, SETTING, METHODS

The study included 1231 couples treated for diverse infertility causes. The fraction of male factor cases was substantial (36.1%). Microinjected oocytes were assessed by a combined TLT-culture system. Oocytes not suitable for TLT assessment, owing to an excess of residual corona cells or inadequate orientation for correct observation, were not analysed. Phenomena relevant to meiotic resumption, pronuclear dynamics, cytoplasmic/cortical modifications, cleavage patterns and embryo quality were annotated and compared between groups.

MAIN RESULTS AND THE ROLE OF CHANCE

Extrusion of the second polar body (PBII) was observed in almost all 2PN/1PN (99.9% and 100.0%, respectively) and in a vast majority of 3PN zygotes (92.1%). Rates of PBII fusion with the ooplasm were much higher in 1PN and 3PN zygotes (P < 0.0001 versus 2PN). The cytoplasmic wave was observed not only in 2PN and 3PN but also in 1PN zygotes (positivity rates of 99.8% and 100% and 82.9%, respectively; P < 0.0001). More rarely, 2PN and 1PN zygotes emitted a third polar body (PBIII). The average times of this event were comparable. The presence and position of the cytoplasmic halo were comparable among the three classes of zygotes. In the 1PN group, the single PN was maternally or paternally derived in 17 and 24 zygotes, respectively, while in the vast majority of 3PN zygotes (121/127) the supernumerary PN was of maternal origin. Average times of maternal PN appearance were comparable, while average times of paternal PN appearance were delayed in 3PN zygotes (P = 0.0127). Compared with the control group, the area of the maternal PN was larger in 1PN zygotes, but smaller in 3PN zygotes (P < 0.0001). The paternal PNs displayed the same trend (P < 0.0001), although such values were consistently smaller than maternal PNs. The area of the third PN in the 3PN group was on average more than 50% smaller than those of maternal and paternal PNs. In maternal PNs of 3PN zygotes, nucleolus precursor bodies (NPBs) aligned along the area of PN juxtaposition at a lower rate compared with the 2PN group. The rate of NPB alignment was ∼50% smaller in 1PN zygotes (P = 0.0001). In paternal PNs, the rates of NPB alignment were not statistically different among the three groups. Asynchronous PN breakdown was increased in 3PN compared with 2PN zygotes (P = 0.0026). In 1PN zygotes, a developmental delay was observed starting from the disappearance of the cytoplasmic halo, reaching 9 h at the time of the first cleavage (P < 0.0001). Higher rates of abnormal cleavage patterns and blastomere fragmentation (P < 0.0001) were observed in 1PN compared to 2N and 3PN zygotes. Cleavage progression was increasingly affected after abnormal fertilization, especially 1PN, finally resulting in blastocyst formation rates of 70.2%, 12.2% and 53.5% in 2PN, 1PN and 3PN embryos, respectively (P < 0.0001). Both maternal and paternal ages were higher in cases involving 3PN fertilization.

LIMITATIONS, REASONS FOR CAUTION

The study data were obtained from ICSI, but not standard IVF, treatments carried out in a single centre. The study findings therefore require independent verification.

WIDER IMPLICATIONS OF THE FINDINGS

This study reports the first detailed morphokinetic map of human abnormal fertilization. Collectively, this evidence prompts new scientific hypotheses and raises clinical questions relevant to the aetiology and the treatment of abnormal fertilization.

STUDY FUNDING/COMPETING INTEREST(S)

This study was supported by the participating institutions. The authors have no conflicts of interest to declare.

TRIAL REGISTRATION NUMBER

N/A.