Related Experiment Video
Updated: May 11, 2025
10:04
A Neural Network-Based Identification of Developmentally Competent or Incompetent Mouse Fully-Grown Oocytes
Published on: March 3, 2018
6.6K
Machine learning center-specific models show improved IVF live birth predictions over US national registry-based
Mylene W M Yao1, Elizabeth T Nguyen2, Matthew G Retzloff3
1R&D Department, Univfy, Los Altos, CA, USA. mylene.yao@univfy.com.
Nature Communications
|April 17, 2025
Summary
Machine learning, center-specific models improve live birth prediction accuracy in fertility care. These models offer better patient counseling and cost-success transparency compared to the SART registry model.
Area of Science:
- Reproductive Medicine
- Biostatistics
- Health Services Research
Background:
- Improving access to in vitro fertilization (IVF) necessitates enhanced patient counseling and cost-effectiveness.
- Accurate live birth prediction (LBP) models are crucial for informed patient decision-making and resource allocation.
- Current LBP models, including the Society for Assisted Reproductive Technology (SART) registry model, may have limitations in precision.
Purpose of the Study:
- To compare the performance of machine learning, center-specific (MLCS) models against the SART national registry model for predicting live birth probability in IVF.
- To evaluate the clinical utility of MLCS models in personalizing prognostic counseling and enhancing cost-success transparency for patients undergoing IVF.
Main Methods:
- A retrospective model validation study was conducted using data from 4635 first-IVF cycles across 6 fertility centers.
- Machine learning, center-specific (MLCS) models were developed and compared to the existing SART registry-based model.
- Model performance was assessed using metrics such as precision-recall area-under-the-curve and F1 score at a 50% LBP threshold.
Main Results:
- MLCS models demonstrated statistically significant improvements in minimizing false positives and negatives compared to the SART model (p < 0.05).
- MLCS models more accurately assigned a higher percentage of patients to LBP thresholds of ≥50% (23% more) and ≥75% (11% more) than the SART model.
- The study found MLCS models to be externally validated, indicating robust performance across different centers.
Conclusions:
- Machine learning, center-specific models offer superior predictive accuracy for live birth probability in IVF compared to the SART model.
- MLCS models enhance clinical utility by enabling more personalized patient counseling and transparent cost-success assessments.
- Further evaluation of MLCS models in a larger, multi-center cohort is recommended to confirm their broader applicability and impact on IVF access.
