Related Experiment Video
Updated: Jun 26, 2025
04:22
Author Spotlight: Advancing Human Cardiac Anatomy Through Multi-Scale Analysis of Hearts
Published on: June 28, 2024
437
Large-scale 3D non-Cartesian coronary MRI reconstruction using distributed memory-efficient physics-guided deep
Chi Zhang1,2, Davide Piccini3,4, Omer Burak Demirel1,2
1Electrical and Computer Engineering, University of Minnesota, 200 Union Street S.E., Minneapolis, MN, 55455, USA.
Magma (New York, N.Y.)
|May 14, 2024
Summary
Physics-guided deep learning (PG-DL) enables high-quality 3D non-Cartesian coronary MRI reconstruction. A novel 2.5D approach improves vessel sharpness and image quality, even with limited training data.
Area of Science:
- Medical Imaging
- Deep Learning
- Cardiovascular MRI
Background:
- Physics-guided deep learning (PG-DL) is a powerful image reconstruction technique.
- Its application to large-scale 3D non-Cartesian MRI is limited by hardware constraints and scarce training data.
Purpose of the Study:
- To enable high-quality PG-DL reconstruction for large-scale 3D non-Cartesian coronary MRI.
- To overcome hardware limitations and limited training data availability.
Main Methods:
- Combined deep learning and MRI reconstruction advances.
- Proposed a 2.5D reconstruction using 2D convolutional neural networks, treating 3D volumes as batches of 2D images.
- Compared 3D and 2.5D PG-DL networks against conventional methods for high-resolution 3D coronary MRI.
Main Results:
- PG-DL reconstructions (3D and 2.5D) outperformed conventional methods quantitatively and qualitatively.
- The 2.5D variant demonstrated superior vessel sharpness and qualitative image quality compared to 3D processing.
Conclusions:
- Achieved high-quality PG-DL reconstruction for large-scale 3D non-Cartesian MRI without compromising image size or network complexity.
- The 2.5D approach enables high-quality reconstruction even with limited training data.
