EFRE IVF 40; AI model for personalized PGTA decision in women Over 40

Maghraby, Hassan; Elmahdy, Mohamed; Gaafar, Sherif; Aboali, Ashraf; Adel, Nehal; ElKassar, Yasser; Hassan, Heba

doi:10.21608/jrme.2025.370139.1037

EFRE IVF 40; AI model for personalized PGTA decision in women Over 40

Document Type : Original Article

Authors

¹ Obstetrics and Gynecology Department, Faculty of Medicine, Alexandria University, Egypt.

² Egyptian Foundation of Reproductive Medicine and Embryology (EFRE), Egypt.

³ Madina Fertility Center, Madina Women's Hospital, Alexandria, Egypt.

⁴ Dar Alteb Infertility Center, Alexandria, Egypt.

10.21608/jrme.2025.370139.1037

Abstract

Background:
Female fertility declines sharply after 40 due to diminished ovarian reserve, compromised oocyte quality, and a significant increase in aneuploidy rates. In vitro fertilization (IVF) strategies, including repeated cycles, embryo pooling, and preimplantation genetic testing for aneuploidy (PGT-A), present complex decision-making challenges requiring personalized, data-driven approaches.
Objective:
This study aims to develop and evaluate AI-driven predictive models using Meta AI to optimize fertility treatment strategies in women over 40, assisting in selecting between repeated IVF cycles and embryo pooling with PGT-A.
Methods:
A generative AI model (Llama 3.2) was employed to extract and synthesize data from peer-reviewed literature. Expert systems were integrated to formulate a structured decision-support framework, leveraging a knowledge base and inference engine. Two predictive models were developed:
1. Pre-Treatment Evaluation Score (PTES) – Assesses overall fertility potential.
2. Embryo Quality Refinement Score (EQRS) – Incorporates embryo-specific factors to refine treatment recommendations.
Results:
Both models demonstrated high predictive accuracy and clinical relevance. Higher scores indicated a preference for embryo pooling with PGT-A, while lower scores supported repeated IVF cycles. Performance metrics, including precision and AUC-ROC, confirmed the models' efficacy in predicting live birth probability.
Conclusion:
AI-driven predictive modeling offers a novel, data-supported approach to personalized fertility care. These models facilitate evidence-based decision-making, potentially improving clinical outcomes. Further validation in real-world clinical settings is warranted.

Keywords