ACMIT and its partners have published a new article titled “An integrated optimization and deep learning pipeline for predicting live birth success in IVF using feature optimization and transformer-based models”.
In vitro fertilization (IVF) has transformed fertility treatment by offering millions of individuals and couples the chance to achieve parenthood. However, predicting the likelihood of a successful outcome—especially live birth—remains a significant clinical challenge due to the complicated relationship among demographic, clinical, and procedural factors. This work presents an innovative AI-based pipeline that is designed to predict live birth success in IVF. The study introduces a hybrid framework that combines Particle Swarm Optimization (PSO) for feature selection with a Tab transformer-based deep learning model. This approach achieved exceptional predictive performance (AUC: 98.4%, Accuracy: 97%) on the large-scale HFEA dataset. In addition to its high accuracy, the model emphasizes clinical transparency through SHAP-based interpretability and demonstrates robustness across various subgroups and preprocessing scenarios. This work marks a promising step toward integrating explainable AI tools into reproductive medicine to support personalized fertility care.
For more information, please visit:
