Related Experiment Video
Updated: Sep 30, 2025
12:29
Magnetic Resonance Imaging Quantification of Pulmonary Perfusion using Calibrated Arterial Spin Labeling
Published on: May 30, 2011
13.9K
Physics-informed neural networks for myocardial perfusion MRI quantification.
Rudolf L M van Herten1, Amedeo Chiribiri2, Marcel Breeuwer3
1Department of Biomedical Engineering, Medical Image Analysis group, Eindhoven University of Technology, Eindhoven, the Netherlands.
Medical Image Analysis
|March 17, 2022
Summary
Physics-informed neural networks (PINNs) improve myocardial perfusion magnetic resonance (MR) quantification by accurately estimating kinetic parameters. This novel approach enhances accuracy in both simulated and real-world patient data.
Area of Science:
- Medical Imaging
- Computational Biology
- Biophysics
Background:
- Dynamic contrast-enhanced magnetic resonance (MR) imaging is crucial for quantifying myocardial perfusion and microvascular function using tracer-kinetic models.
- Challenges in traditional model fitting include low signal-to-noise ratio and limited temporal resolution, leading to inaccurate kinetic parameter estimates.
- Multi-compartment exchange models offer physiological plausibility but are sensitive to data quality limitations.
Purpose of the Study:
- To introduce and validate a framework using physics-informed neural networks (PINNs) for myocardial perfusion MR quantification.
- To enhance the accuracy and reliability of kinetic parameter estimation in myocardial perfusion imaging.
- To provide a versatile computational scheme for inferring physiological parameters from perfusion MR data.
Main Methods:
- Implementation of a physics-informed neural network (PINN) framework for myocardial perfusion MR.
- Training neural networks to fit observed perfusion MR data while enforcing physical conservation laws from multi-compartment exchange models.
- Validation of the PINN approach using both in silico (simulated) and in vivo (patient) datasets.
Main Results:
- In silico validation showed a significant reduction in mean-squared error compared to standard non-linear least squares fitting.
- In vivo studies demonstrated that PINN-derived kinetic parameters are comparable to established literature values.
- Generated parameter maps from the PINN method aligned with clinical diagnoses in patient studies.
Conclusions:
- Physics-informed neural networks offer a robust and accurate method for myocardial perfusion MR quantification.
- PINNs overcome limitations of traditional fitting methods, improving the reliability of kinetic parameter estimation.
- This framework shows promise for enhanced clinical diagnosis and understanding of myocardial perfusion.
