From Statistical Regression to Explainable AI: A Synergistic Approach for Predicting Euploid Embryo Yield Based on Maternal Age, AMH, and a Visual Decision-Support System

Abstract

Background: Predicting the number of euploid embryos is critical for optimising IVF outcomes and managing patient expectations. While maternal age and anti-Müllerian hormone (AMH) are established markers of ovarian reserve, their combined predictive power regarding chromosomal normality remains a subject of clinical debate. Artificial intelligence is increasingly being explored in assisted reproduction as a non-invasive, data-driven approach to estimate embryo ploidy. By leveraging advanced models such as convolutional neural networks (CNNs) and machine learning algorithms to evaluate morphological and morphokinetic characteristics from time-lapse sequences, AI contributes to improving the accuracy and objectivity of embryo selection. Objective: This study evaluated the statistical association between maternal age, AMH levels, and fertilisation methods (IVF, ICSI, IMSI) and euploid embryo yield. A secondary objective was to translate these clinical findings into a visual decision-support system (DSS) grounded in an Explainable AI (XAI) framework. Methods: A retrospective observational study was conducted on 31 patients undergoing IVF with PGT-A. Statistical significance was assessed using one-way ANOVA and multiple linear regression. Building on these data, a specialised decision-support system was developed using React 19 and TypeScript, employing a binomial probability model to translate clinical biomarkers into intuitive success simulations.
Results: Patients younger than 35 years exhibited significantly higher AMH levels (p = 0.033) and a higher mean number of euploid embryos (p = 0.032) compared to those greater than 35. The fertilisation method did not significantly influence euploidy outcomes (p = 0.990). The regression model was statistically significant (p = 0.030), explaining 22.1% of the variance. However, none of the individual predictors reached statistical significance, suggesting that the observed effect may be driven by the combined contribution of the variables rather than by independent effects. The resulting DSS operationalises these findings in a preliminary manner through real-time attrition modelling and "Opportunity Cost" visualisations. Conclusion: Maternal age may represent an important factor in embryo euploidy, while AMH provides a quantitative baseline for embryo yield. By synergising retrospective data with explainable AI, the developed framework offers a transparent, data-driven approach to fertility counselling. This study indicates that integrating statistical analysis with a visual decision-support system effectively bridges the gap between raw clinical data and patient-centred practice, facilitating more objective decision-making in in vitro fertilisation.The development of a conceptual decision support system based on these findings derived a secondary objective of the paper by exploring it.