Related Experiment Video
Updated: Jun 13, 2025
06:56
Human Fetal Blood Flow Quantification with Magnetic Resonance Imaging and Motion Compensation
Published on: January 7, 2021
2.4K
Automatic flow planning for fetal cardiovascular magnetic resonance imaging
Sara Neves Silva1, Tomas Woodgate2, Sarah McElroy3
1Research Department for Early Life Imaging, School of Biomedical Engineering & Imaging Sciences, King's College London, London, UK; Research Department for Medical Engineering, School of Biomedical Engineering & Imaging Sciences, King's College London, London, UK.
Summary
Automated fetal flow imaging planning using deep learning (OWL) enhances access to cardiovascular magnetic resonance (CMR) by simplifying real-time 2D phase-contrast sequence setup.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Cardiovascular Imaging
Background:
- Fetal flow imaging, specifically 2D phase-contrast (PC) imaging, is crucial for assessing fetal cardiovascular health.
- Current manual planning for PC imaging is time-consuming and requires specialized expertise, limiting its widespread use.
- Automating the planning process can significantly widen access to this vital diagnostic tool.
Purpose of the Study:
- To develop and evaluate an automated system (OWL) for real-time planning of fetal 2D phase-contrast flow imaging.
- To assess the feasibility and accuracy of deep learning networks for fetal body localization and cardiac landmark detection.
- To compare the performance of automated planning with manual planning in terms of accuracy and efficiency.
Main Methods:
- Two deep learning networks were trained for fetal body localization and cardiac landmark detection on coronal whole-uterus scans.
- The system was implemented for real-time automatic planning of phase-contrast sequences.
- Evaluation involved retrospective analysis of 10 datasets and prospective assessment in 7 fetal subjects, comparing automated vs. manual planning.
Main Results:
- The automated system (OWL) was successfully implemented in 6 out of 7 prospective cases.
- High accuracy was achieved in fetal body localization (Dice score 0.94±0.05) and cardiac landmark detection (e.g., descending aorta 5.77±2.91 mm).
- Indexed flow measurements showed minimal difference (-1.8%) between automated and manual planning, with comparable planning quality.
Conclusions:
- Automated planning of 2D phase-contrast cardiovascular magnetic resonance (CMR) for major fetal vessels is feasible using OWL.
- The system demonstrated successful real-time application at 0.55T, suggesting potential generalization across field strengths.
- This technology has the potential to extend access to advanced fetal cardiovascular imaging beyond specialized centers.
