Related Experiment Video
Updated: Jul 8, 2026
06:49
A Semi-high-throughput Imaging Method and Data Visualization Toolkit to Analyze C. elegans Embryonic Development
Published on: October 29, 2019
6.6K
Deep learning classification integrating embryo images with associated clinical information from ART cycles
Mohamed Salih1, Christopher Austin2, Krishna Mantravadi3
1Department of Obstetrics and Gynaecology, Monash University, 246 Clayton Road, Clayton, VIC, 3168, Australia.
Scientific Reports
|May 21, 2025
Summary
Artificial Intelligence (AI) enhances IVF success prediction by integrating patient data and embryo images. A fused AI model combining clinical information and blastocyst images achieved the highest accuracy in predicting pregnancy outcomes.
Area of Science:
- Reproductive Medicine
- Artificial Intelligence
- Computer Vision
Background:
- Predicting clinical pregnancy outcomes in single embryo transfer (SET) remains challenging.
- Existing methods often rely solely on clinical data or static embryo images.
- Advanced AI offers potential for improved predictive accuracy by integrating diverse data sources.
Purpose of the Study:
- To develop and evaluate an advanced Artificial Intelligence (AI) model for predicting clinical pregnancy and live birth outcomes in SET procedures.
- To compare the predictive performance of AI models using clinical data, blastocyst images, and a fusion of both.
- To identify key clinical and embryonic features influencing pregnancy prediction through AI model visualization.
Main Methods:
- Developed three AI models: a Clinical Multi-Layer Perceptron (MLP) for patient data, an Image Convolutional Neural Network (CNN) for blastocyst images, and a Fusion model combining both.
- Trained and tested models on 1503 international IVF treatment cycles.
- Evaluated model performance using accuracy, average precision, and Area Under the Curve (AUC).
- Utilized a visualization process to determine feature importance.
Main Results:
- The Clinical MLP model achieved 81.76% accuracy, 90% average precision, and 0.91 AUC.
- The Image CNN model achieved 66.89% accuracy, 74% average precision, and 0.73 AUC.
- The Fusion model demonstrated superior performance with 82.42% accuracy, 91% average precision, and 0.91 AUC.
- Female and male age were identified as crucial clinical factors, while Trophectoderm was the most significant blastocyst feature.
Conclusions:
- Integrating patient clinical information with blastocyst images via a fused AI model significantly enhances the prediction of IVF clinical pregnancy outcomes.
- AI models, particularly the fusion approach, offer a powerful tool for improving IVF success rates by providing more informed predictions.
- The study highlights the potential of AI in reproductive medicine to optimize treatment strategies and patient care.
