Artificial intelligence in the in vitro fertilization laboratory: a committee opinion

Artificial intelligence (AI) became more mainstream in the 1990s as chess champions were defeated by Deep Blue, the IBM chess-playing supercomputer. In the 2000s, AI became a prevalent part of discussions in medicine, with increased use in diagnosing diseases by analyzing medical images and data. It has now reached into the field of reproductive medicine, and like many other fields, there are questions regarding its true utility and implementation. The current practice committee document aims to highlight areas where AI could be a valuable adjunct in the IVF laboratory, address the current evidence for its utility, and outline expectations on necessary validation.

DEVELOPING AI MODELS

There are several key steps one must consider while developing an AI-based processing algorithm. For example, AI-based embryo development models rely on datasets developed from time-lapse imaging (TLI) that may include embryo images captured at different stages, along with the timings of key developmental events. The development of AI algorithms first requires the dataset to be prepared by organizing the video data and associated information on the desired outcome (e.g., prediction of blastocyst formation, euploidy, clinical pregnancy, implantation). Relevant features must next be extracted from raw data, images, and/or video data, capturing morphological characteristics, temporal dynamics, and other important information using computer vision techniques. The third step requires the selection of a machine learning model for analyzing time-lapse videos. This is followed by training the learning algorithm that adjusts the model using high-quality data to minimize error. Subjective labeling refers to the process of assigning labels or categories to data on the basis of individual judgment or interpretation (for example, manual embryo development annotations, embryo grading). Because it can introduce bias and inconsistency, subjective labeling can lead to variations in the dataset in many cases and can affect the accuracy and reliability of the learning algorithm. Similarly, training on an unbalanced dataset (meaning that the proportions of samples in different classes or categories are heavily skewed, with one class dominating the dataset) can be problematic and may not accurately represent the real-world distribution of outcomes. Validation studies should be conducted using new data to assess the model’s performance. Finally, the AI model must be tested in a real-life scenario, preferably a randomized controlled trial (RCT), where the flaws and performance can be observed in a clinical setting.

IMPACT OF AI ON THE ROLE OF EMBRYOLOGISTS IN THE UCD LABORATORY

The impact of AI in the in vitro fertilization (IVF) laboratory can be considered at different levels. First, although less clinical in some sense, AI can impact logistics. This can range from an impact on scheduling caseloads to assisting in tasks that would save time for the embryologist. The metrics that need to be applied in terms of logistics are twofold and can encompass savings in time and expense. These should be assessed individually by clinics, because the number of cases and laboratory personnel needed will depend on how impactful AI may be in single clinics that differ in size. For example, the ability of AI to predict stimulation response (1), in addition to being an interesting adjunct to optimize the dosage and timing of medications, may also be useful in predicting or adjusting how many cases will be retrieved the following week. A larger clinic could potentially homogenize the number of retrievals per day or lessen loads on weekends (2, 3), potentially aiding in coordinating staffing levels. Second, tasks that the embryologists perform may be assisted by AI. The main focus has centered on automating embryo grading; however, other tasks, including witnessing (4, 5) and storage management of embryos (6), are starting to be influenced by AI. In this scenario, AI may also facilitate the automation of various techniques, including intracytoplasmic sperm injection (ICSI) and sperm selection (7, 8). Regardless, the true clinical impact is yet to be validated, and many of these applications remain to be robustly tested (9).

EMBRYO GRADING BY AI

Globally, assisted reproductive technology (ART) has become more standardized in how embryology is practiced, and mandated reporting in many countries has led to the development of standardized embryo grading systems (10–12). Homogeneous standard embryo grading systems (13) enable scientists to compare data from international study sites while creating a language when describing embryos that can be understood globally, both by IVF professionals and patients.

Currently, three primary grading systems exist for blastocysts, whereas multiple systems exist for cleavage stage embryos, each with modifications in some laboratories to make them more robust. The most widely used blastocyst grading systems in the US are the Gardner Grading System (GGS), the Veeck Grading System, and the Society for Assisted Reproductive Technology (SART) grading system. The details of each of these systems are outside the scope of this article and are described elsewhere (12–15). All of these are performed manually by trained embryologists, and encompass an expansion rating, followed by a quality score for the inner cell mass (ICM) and trophectoderm (TE), using numbers and/or alphabetical letters of descriptions (Good, Fair, Poor). It is universally accepted that the lower the number of the assigned score (letter), the higher the morphological quality of the cell population described, whereas the expansion numerical grade increases in value. These are universally accepted scales that can be correlated to one another and allow embryo records created in one laboratory to be interpreted in another.

Advantages of AI technology for embryo grading, at the moment, are purely to benefit workflow in the laboratory. This is particularly applicable when using time-lapse and was clearly shown in a recent RCT by Illingworth et al (16), where the use of an AI model provided an almost 10-fold reduction in the time required for evaluation of blastocysts cultured in the Embryoscope (Vitrolife, Sweden) time-lapse incubator. The implementation of a similar model may allow laboratory staff to grade embryos at a convenient time, or better yet, defer grading to the AI model. It may also remove pressure from the embryologist to make a subjective call on a grade, allowing better consistency and homogeneity in embryo evaluation. This technology is still in its infancy, and although several companies are racing to release products, there is still a need for rigorous evaluation (9).

Studies examining concordance rates of embryo grading between embryologists have shown a broad variability in grading (17, 18). This is also evident when examining the assessment of embryos using the American Society for Reproductive Medicine (ASRM) EDGE tool (https://www. asrm.org/asrm-academy/asrm-academy-on-the-go/embryo-data-grading–evaluation/). The concordance and discordance among embryologists in ranking embryos were more pronounced when compared with the rankings of blastocysts generated by available AI models (18). The application of AI and image recognition has the ability to enhance the reliability and provide higher consistency during the process of embryo selection and disposition. It would also allow greater dependency when comparing IVF laboratories because embryo grading will be more standardized. The evidence of the ability to use AI to assist in embryo grading is promising (16), but further studies are required to validate its use in the IVF laboratory.

STATIC EMBRYO IMAGES AND AI

Another AI model approach (contrary to TLI use) has been to rely on static images of embryos, mainly blastocysts. In the rapidly evolving domain of AI, recognition software interrogating static images provides a vital testing ground for exploring, learning, and refining algorithms and models. Static images present several complexities to AI systems that may not be apparent to the human eye. Unlike dynamic or moving images, they encapsulate a single moment frozen in time. They are a collection of pixels, each carrying unique information about color, intensity, and texture. Interpreting these pixels and making sense of the entire composition is a daunting task for an AI system.

In healthcare, AI's ability to analyze medical images like CT scans or roentgenograms improves diagnostic capabilities (19), thereby enabling early detection of diseases and accelerating personalized care. In reproduction, a number of companies have taken the approach of static image analysis of blastocysts, recognizing that not all clinics have adopted TLI. A study by Loewke et al (20) relied on ranking embryos with a score after static image assessment. They concluded that there was excellent potential for AI to rank blastocyststage embryos, but highlighted limitations related to image quality, bias, and granularity of the ranking scores. VerMilyea et al (21) and Diakiw et al (22) have also shown that static images may be used to standardize blastocyst grading, embryo potential, and possibly time to pregnancy. As mentioned previously, these studies have again been performed on retrospective datasets. The question still remains whether the single static image should be taken at a specific developmental time point.

In contrast to using the blastocyst as a time point of imaging, others are focusing on the oocyte. Recently, Hall et al (23) have created a deep learning algorithm for oocyte grading to predict blastocyst formation and ploidy, on the basis of retrospective image analyses and a prospective clinical study. Numerous other investigations are ongoing as an objective measure of oocyte developmental competence that may have implications for elective oocyte cryopreservation.

TIME LAPSE IMAGING AND AI

The current research efforts in TLI and AI in human IVF aim largely to improve implantation rates, minimize the risk of multiple pregnancy, reduce miscarriage rates, and enhance the detection of euploid embryos. The greatest advantage associated with TLI stems from the theoretical benefit of providing better solutions and standardization for selecting which embryo may have the best potential for establishing a pregnancy. This is potentially akin to having the laboratories’ most experienced embryologist perform grading every time. Single embryo transfer has driven the desire to optimize embryo selection techniques. Time lapse imaging has allowed the ability to track the embryos' intricate details (multinucleation, fragmentation, direct and asymmetric divisions) and the timing of morphological changes, as well as measuring cell-cycle lengths during embryo development. In addition, TLI has generated a substantial amount of morphokinetic data, leading to the development of embryo selection algorithms (ESAs) as potential predictors of IVF outcomes. Despite efforts to develop universal ESAs, validation in randomized trials has been limited, but those that have been performed have largely failed to show benefit when pitting AI-assisted embryo selection against standard morphology assessment (16, 24–26). One reason may be that each algorithm has been developed using specific sets of embryos from different IVF clinics, which were derived from diverse patient populations with differences in their clinical and laboratory practices (protocols). The algorithms and their contents remain a black box. As a result, when these algorithms are applied retrospectively or prospectively to different sets of embryos, their predictive capacity does not allow them to yield consistent results. Another reason could be that blastocyst morphology grading itself is an excellent test of viability already (27).

Automated embryo grading

A significant disadvantage of ESAs is the requirement of manual annotation of morphological features and morphokinetic data for each embryo. These annotations can be employed as an input for statistical or machine learning-based scoring tools. However, one major challenge is that the annotation and grading of time-lapse data are a subjective process that exhibits inherent variability among different embryologists and even within the same embryologist (28). This subjective nature of the process has the potential to directly affect the prediction capability of the ESAs. Although advanced TLSs provide AI-powered auto-annotation capability, they still hold the burden of allocating a significant amount of embryologists’ time to complete and verify the annotations by ESAs. The annotation of embryo grades by AI has advanced, and platforms that automatically assign grades exist from several AI companies. In a systematic review, TLI and convolutional neural networks (CNN) models' diagnostic test accuracy was evaluated (29). The systematic review identified 22 retrospective studies for the evaluation. These studies analyzed a total of 522,516 images of 222,998 embryos and evaluated outcomes such as, successful IVF, blastocyst stage classification, and blastocyst quality. Most studies reported an accuracy rate of over 80%, with some AI models outperforming embryologists.

Prediction of embryo potential

Compared with ESAs, AI-based technologies have emerged as helpful tools for evaluating embryo development using TLIs. Particularly, deep learning algorithms can analyze the entire raw time-lapse video data without the need for manually annotated parameters. By leveraging every data point collected from the time-lapse videos, deep learning algorithms have been developed to presumably predict the probability of achieving a successful full-term pregnancy or rank embryos in order of transfer preference. This capability, if proven, could improve ART outcomes and provide more precise and personalized fertility treatments (30).

There are numerous examples of AI implementation in the TLI literature, the majority being with retrospective datasets. For example, a deep learning model called IVY (31) was developed and trained using time-lapse videos to predict the probability of pregnancy with fetal heart activity directly from the videos, without manual annotation or morphological assessment. The model demonstrated a high predictive power with an average area under the curve (AUC) of 0.93 (31). In a recent study, researchers explored a hybrid learning model that incorporated video analysis and clinical features (32). The study initially involved constructing a CNN with a ResNet backbone architecture using PyTorch for TLI processing. Later, the video score obtained from the CNN was combined with preprocessed multicentral clinical data by using a machine learning (ML) model (XGBoost). This combined approach resulted in a sevenfold improvement in the prediction of pregnancy, as measured by the average AUC for fetal heart assessment.

More importantly, the largest RCT (16) has recently been published and was unable to demonstrate the noninferiority of deep learning for clinical pregnancy rate when compared with standard morphology and a predefined prioritization scheme. In the Illingworth et al (16) study, a deep learning algorithm AI model ‟iDAScore version 1’’ was trained and evaluated on the basis of a large dataset from 18 IVF centers consisting of 115,832 embryos, of which 14,644 embryos were transferred embryos with known outcome. This algorithm was then tested through the RCT by comparing embryos assessed using standard morphology criteria (Gardner grade) for the control arm, or iDAScore for the study arm. The iDAScore group exhibited a clinical pregnancy rate of 46.5% (248 of 533 patients), compared with 48.2% (257 of 533 patients) in the control arm (risk difference: –1.7%, 95% CI: –7.7 to 4.3, P = .62).

Prediction of embryo ploidy status

A further aim has been to predict euploidy in blastocysts. The largest study to date reached predictive accuracies in the 60%–75% range (22). The investigators proposed that this shows future potential as a standardized supplementation to traditional methods of embryo selection. The AI prediction models have now been published that fully automate the prediction of euploidy to an AUC of 0.76 (33). For patients desiring preimplantation genetic testing for aneuploidy (PGT-A), an AI-PGT algorithm could eventually provide an alternative.

In conclusion, the use of artificial intelligence in time-lapse monitoring has the potential to improve embryo evaluation by providing more efficient, accurate, and objective assessments. The AI models focused on blastocyst stage classification demonstrated the best predictions.

VALIDATION OF AI IN THE IVF LABORATORY–THE NEED FOR RCTs

We have seen with other technologies that their adoption in the field of IVF has led to controversy (9). Add-ons are the best example of this (34), whereby many have not been shown to improve the live birth rate from IVF, yet are entrenched in many IVF practices. Furthermore, caution should be practiced in relation to commercial pressures related to the adoption of AI from AI companies themselves and by clinics portraying unproven benefits of AI in an attempt to gain a commercial advantage. Enthusiasm about adopting new technologies should be balanced with the need for evidence-based studies looking at patient outcomes, safety issues, and improvement in laboratory efficiency. Randomized trials are critical to demonstrate the effectiveness of these benefits, in particular, using AI for embryo selection to improve pregnancy outcome and/or time to live birth. The recent RCT by Illingworth et al (16) was not able to demonstrate noninferiority using a deep learning algorithm for the clinical pregnancy rate when compared with standard morphology. The study, however, just as importantly, showed no harm in using AI selection vs. human selection. Other laboratory-based applications of AI would also need to demonstrate that they are improving time, quality, and safety for IVF laboratory procedures, as long as clinical outcomes are not compromised.

SUMMARY AND CONCLUSION

The implementation of AI in the IVF laboratory is still at an early stage, and the recommendation is to proceed with caution. A number of complex questions, independent of any clinical benefits, are associated with the adoption of AI, including its blackbox nature (35), data ownership, and regulation by the Food and Drug Administration (FDA) or equivalent bodies. For the IVF laboratory, the use of AI for embryo selection could be of huge benefit as an adjunct technology to aid selection of which blastocyst is likely to implant. Whereas numerous retrospective analyses have been published indicating a benefit of AI in selecting the most viable embryo for transfer, appropriately designed RCTs, or alternative methods, are imperative to evaluate the risks and benefits of incorporating AI in IVF laboratory processes before adoption into clinical practice. One large RCT has so far not been able to demonstrate noninferiority when using an algorithm to perform embryo selection compared with standard morphology to improve clinical pregnancy rates. More well-designed prospective studies, including RCTs, examining the validity of AI models in improving outcomes related to live birth, time, safety, reduction of errors, and cost are needed before widespread adoption.

Acknowledgments

This report was developed under the direction of the Practice Committee of the American Society for Reproductive Medicine (ASRM) as a service to its members and other practicing clinicians. Although this document reflects appropriate management of a problem encountered in the practice of reproductive medicine, it is not intended to be the only approved standard of practice or to dictate an exclusive course of treatment. Other plans of management may be appropriate, taking into account the needs of the individual patient, available resources, and institutional or clinical practice limitations. The Practice Committee and the Board of Directors of the American Society for Reproductive Medicine have approved this report. This document was reviewed by ASRM members, and their input was considered in the preparation of the final document. The following members of the ASRM Practice Committee participated in the development of this document: Clarisa Gracia, M.D., M.S.C.E.; Rebecca Flyckt, M.D.; Denny Sakkas, Ph.D.; Karl Hansen, M.D., Ph.D.; Tarun Jain, M.D.; Suleena Kalra, M.D., M.S.C.E.; Bruce Pier, M.D.; Belinda Yauger, M.D.; Torie C. Plowden, M.D., M.P.H.; Ryan Smith, M.D.; Mark Trolice, M.D., M.B.A.; Suneeta Senapati, M.D.; Robert Brannigan, M.D.; Amy Sparks, Ph.D., H.C.L.D; Jared Robins, M.D.; Chevis N Shannon, Dr.Ph., M.B.A., M.P.H.; Jessica Goldstein, R.N.; and Madeline Brooks, M.B.A., M.P.H. The Practice Committee also acknowledges the special contribution of Denny Sakkas, Ph.D.; Charles Bormann, Ph.D.; Cihan Halicigil, Ph.D., M.S., H.C.L.D.; Sangita Jindal Ph.D., H.C.L.D.; Liesl Nel-Thermaat Ph.D., H.C.L.D., M.B.A.; Salustiano Ribeiro, B.S., M.S.; Mitchel Schiewe, Ph.D., M.S.; and Nikica Zaninovic, Ph.D., M.S.; in the preparation of this document. All committee members disclosed commercial and financial relationships with manufacturers or distributors of goods or services used to treat patients. Members of the Committee who were found to have conflicts of interest on the basis of the relationships disclosed did not participate in the discussion or development of this document.

REFERENCES

1. Hariton E, Pavlovic Z, Fanton M, Jiang VS. Applications of artificial intelligence in ovarian stimulation: a tool for improving efficiency and outcomes. Fertil Steril 2023;120:8–16.
2. Youngster M, Reuvenny S, Luz A, Hourvitz R, Moran E, Ben-Mayor Bashi T, et al. Optimizing workload balance using artificial intelligence. Fertil Steril 2024;122:178–80.
3. Letterie G, Mac Donald A. Artificial intelligence in in vitro fertilization: a computer decision support system for day-to-day management of ovarian stimulation during in vitro fertilization. Fertil Steril 2020;114:1026–31.
4. Sterckx J, Wouters K, Mateizel I, Segers I, De Vos A, Van Landuyt L, et al. Electronic witnessing in the medically assisted reproduction laboratory: insights and considerations after 10 years of use. Hum Reprod 2023;38: 1529–37.
5. Hammer KC, Jiang VS, Kanakasabapathy MK, Thirumalaraju P, Kandula H, Dimitriadis I, et al. Using artificial intelligence to avoid human error in identifying embryos: a retrospective cohort study. J Assist Reprod Genet 2022; 39:2343–8.
6. Go KJ, Hudson C. Deep technology for the optimization of cryostorage. J Assist Reprod Genet 2023;40:1829–34.
7. Jiang VS, Kartik D, Thirumalaraju P, Kandula H, Kanakasabapathy MK, Souter I, et al. Advancements in the future of automating micromanipulation techniques in the IVF laboratory using deep convolutional neural networks. J Assist Reprod Genet 2023;40:251–7.
8. Costa-Borges N, Munne S, Alb oE, Mas S, Castell oC, Giralt G, et al. First babies conceived with automated intracytoplasmic sperm injection. Reprod Biomed Online 2023;47:103237.
9. ASRM. Moving innovation to practice: an Ethics Committee opinion. Available at: https://www.asrm.org/practice-guidance/ethics-opinions/moving-innovation-to-practice-an-ethics-committee-opinion-2021/?_t_ tags=siteid:01216f06-3dc9-4ac9-96da-555740dd020c,language:en&_t_ hit.id=ASRM_Models_Pages_ContentPage/_59573c13-d570-4f00-bef9-
   c4e9003cbf62_en&_t_hit.pos=19. Accessed June 1, 2026.
10. Alpha Scientists in Reproductive Medicine and ESHRE Special Interest Group of Embryology. The Istanbul consensus workshop on embryo assessment: proceedings of an expert meeting. Hum Reprod 2011;26:1270–83.
11. ALPHA Scientists In Reproductive Medicine; ESHRE Special Interest Group Embryology. Istanbul consensus workshop on embryo assessment: proceedings of an expert meeting. Reprod Biomed Online 2011; 22:632–46.
12. Racowsky C, Vernon M, Mayer J, Ball GD, Behr B, Pomeroy KO, et al. Standardization of grading embryo morphology. Fertil Steril 2010;27: 437–9.
13. Working Group on the update of the ESHRE/ALPHA Istanbul Consensus, Coticchio G, Ahlstrom A, Arroyo G, Balaban B, Campbell A, et al. The Istanbul consensus update: a revised ESHRE/ALPHA consensus on oocyte and embryo static and dynamic morphological assessment. Hum Reprod 2025;40:989–1035.
14. Gardner DK, Schoolcraft WB. Culture and transfer of human blastocysts. Curr Opin Obstet Gynecol 1999;11:307–11.
15. Veeck LL, Zaninovic N. Grading criteria for human blastocysts, An Atlas of Human Blastocysts. 1st. New York: Parthenon Publishing; 2003:118.
16. Illingworth PJ, Venetis C, Gardner DK, Nelson SM, Berntsen J, Larman MG, et al. Deep learning versus manual morphology-based embryo selection in IVF: a randomized, double-blind noninferiority trial. Nat Med 2024;30: 3114–20.
17. Bormann CL, Thirumalaraju P, Kanakasabapathy MK, Kandula H, Souter I, Dimitriadis I, et al. Consistency and objectivity of automated embryo assessments using deep neural networks. Fertil Steril 2020;113:781–7.e1.
18. Zaninovic N, Sierra JT, Malmsten JE, Rosenwaks Z. Embryo ranking agreement between embryologists and artificial intelligence algorithms. F S Sci 2024;5:50–7.
19. Seah JCY, Tang JSN, Tran A. Drafting the future: the dawn of AI report generation in radiology. Radiology 2025;316:e243378.
20. Loewke K, Cho JH, Brumar CD, Maeder-York P, Barash O, Malmsten JE, et al. Characterization of an artificial intelligence model for ranking static images of blastocyst stage embryos. Fertil Steril 2022;117:528–35.
21. Vermilyea M, Hall JMM, Diakiw SM, Johnston A, Nguyen T, Perugini D, et al. Development of an artificial intelligence-based assessment model for prediction of embryo viability using static images captured by optical light microscopy during IVF. Hum Reprod 2020;35:770–84.
22. Diakiw SM, Hall JMM, Vermilyea M, Lim AYX, Quangkananurug W, Chanchamroen S, et al. An artificial intelligence model correlated with morphological and genetic features of blastocyst quality improves ranking of viable embryos. Reprod Biomed Online 2022;45:1105–17.
23. Hall JMM, Nguyen TV, Dinsmore AW, Perugini D, Perugini M, Fukunaga N, et al. Use of federated learning to develop an artificial intelligence model predicting usable blastocyst formation from pre-ICSI oocyte images. Reprod Biomed Online 2024;49:104403.
24. AhlstromA, Lundin K, Lind AK, Gunnarsson K, Westlander G, Park H, et al. A double-blind randomized controlled trial investigating a time-lapse algorithm for selecting Day 5 blastocysts for transfer. Hum Reprod 2022;37: 708–17.
25. Bhide P, Chan DYL, Lanz D, Alqawasmeh O, Barry E, Baxter D, et al. Clinical effectiveness and safety of time-lapse imaging systems for embryo incubation and selection in in-vitro fertilisation treatment (TILT): a multicentre, three-parallel-group, double-blind, randomised controlled trial. Lancet 2024;404:256–65.
26. Kieslinger DC, Vergouw CG, Ramos L, Arends B, Curfs M, Slappendel E, et al. Clinical outcomes of uninterrupted embryo culture with or without time-lapse-based embryo selection versus interrupted standard culture (SelecTIMO): a three-armed, multicentre, double-blind, randomised controlled trial. Lancet 2023;401:1438–46.
27. Sakkas D. The 'golden fleece of embryology' eludes us once again: a recent RCT using artificial intelligence reveals again that blastocyst morphology re-
    mains the standard to beat. Hum Reprod 2025;40:4–8.
28. Storr A, Venetis CA, Cooke S, Kilani S, Ledger W. Inter-observer and intraobserver agreement between embryologists during selection of a single day 5 embryo for transfer: a multicenter study. Hum Reprod 2017;32:307–14.
29. Berman A, Anteby R, Efros O, Klang E, Soffer S. Deep learning for embryo evaluation using time-lapse: a systematic review of diagnostic test accuracy. Am J Obstet Gynecol 2023;229:490–501.
30. Dimitriadis I, Zaninovic N, Badiola AC, Bormann CL. Artificial intelligence in the embryology laboratory: a review. Reprod Biomed Online 2022;44:435–48.
31. Tran D, Cooke S, Illingworth PJ, Gardner DK. Deep learning as a predictive tool for fetal heart pregnancy following time-lapse incubation and blastocyst transfer. Hum Reprod 2019;34:1011–8.
32. Duval A, Nogueira D, Dissler N, Maskani Filali M, Delestro Matos F, Chansel-Debordeaux L, et al. A hybrid artificial intelligence model leverages multi-centric clinical data to improve fetal heart rate pregnancy prediction across time-lapse systems. Hum Reprod 2023;38: 596–608.
33. Rajendran S, Brendel M, Barnes J, Zhan Q, Malmsten JE, Zisimopoulos P, et al. Automatic ploidy prediction and quality assessment of human blastocysts using time-lapse imaging. Nat Commun 2024;15:7756.
34. ESHRE Add-ons working group, Lundin K, Bentzen JG, Bozdag G, Ebner T, Harper J, et al. Good practice recommendations on add-ons in reproductive medicine † . Hum Reprod 2023;38:2062–104.
35. Afnan MAM, Liu Y, Conitzer V, Rudin C, Mishra A, Savulescu J, et al. Interpretable, not black-box, artificial intelligence should be used for embryo selection. Hum Reprod Open 2021;2021:hoab040.