Innovative AI models for clinical decision-making: predicting blastocyst formation and quality from time-lapse embryo images up to embryonic day 3

Abstract Accurate embryo assessment on embryonic day 3 of assisted reproductive technology (ART) is crucial for deciding whether to continue the culture until day 5 (blastocyst stage) or opt for earlier transfer or cryopreservation. Prolonged culture often improves pregnancy outcomes in patients with multiple high-quality embryos, but may offer limited benefits for older patients or those with few available embryos. To address this clinical challenge, analyzing embryo quality in early stages by artificial intelligence (AI) can be useful. We retrospectively analyzed images of 7,111 two-pronuclear embryos (Veeck grade ≤3) using four different time-lapse incubators. We fine-tuned ImageNet-1k-pretrained NASNet-A Large to automatically classify each time-lapse image into 17 morphological categories, including cell stages and Veeck grades 1–3. This model achieved 95% cell-stage accuracy on the test set. We combined these annotations with age at egg retrieval in a gradient boosting framework (XGBoost) to predict blastocyst formation, good blastocysts, and poor blastocyst + arrested embryos (PBAE). The ROC AUCs were 0.87, 0.88, and 0.87 for blastocyst formation, good blastocysts, and PBAE, respectively, indicating good predictive performance for day 3 embryo assessment. Notably, the PBAE model reached a precision-recall AUC of 0.90, accurately identifying embryos unlikely to benefit from extended culture. This revolutionary AI prediction model could ensure transparency and addresses the “black box” limitation often associated with AI. By integrating a high-accuracy auto-annotation pipeline with interpretable AI (via SHapley Additive exPlanations), our device-independent approach supports appropriate embryo-specific decisions, potentially reducing unnecessary culture, optimizing workflows, and improving clinical outcomes in ART. Capsule Artificial intelligence models accurately predicted blastocyst formation and quality using time-lapse images and age on embryonic day 3, supporting clinical decision-making regarding blastocyst culture or early embryo transfer. Highlights Time-lapse (24–64 hpi) enables accurate day 3-based blastocyst prediction (AUC=0.87). Poor blastocyst + arrested embryo model offers PR AUC=0.90 for early decisions. Robust across four incubator types and age groups, demonstrating broad utility. SHAP analysis clarifies how morphological features drive AI predictions. Embryo-specific decisions reduce unnecessary culture and improve patient outcomes. Competing Interest Statement The authors have declared no competing interest. Funding Statement This work was supported by JSPS KAKENHI grant number JP23K08821. Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: This study was conducted in accordance with the principles embodied in the Declaration of Helsinki and was approved by the Ethics Committee of Kyoto University Graduate School and Faculty of Medicine (approval number R3320). Informed consent was obtained in the form of opt-out through the websites of facilities. I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes Footnotes We updated the abstract and references to conform to the new journal's guidelines and made minor textual refinements throughout, ensuring clarity and consistency without altering the study's overall content.