Related Experiment Video
Updated: Jul 11, 2025
11:13
Quantification of Mouse Heart Left Ventricular Function, Myocardial Strain, and Hemodynamic Forces by Cardiovascular Magnetic Resonance Imaging
Published on: May 24, 2021
6.4K
High-resolution spiral real-time cardiac cine imaging with deep learning-based rapid image reconstruction and
Junyu Wang1, Marina Awad2, Ruixi Zhou3
1Department of Medicine, Cardiovascular Medicine, Stanford University, Stanford, California, USA.
NMR in Biomedicine
|November 5, 2023
Summary
Deep learning reconstruction (DESIRE) and segmentation significantly accelerate cardiac MRI by enabling whole-heart reconstruction in one minute. This advanced technique accurately quantifies left ventricular ejection fraction, matching manual segmentation performance.
Area of Science:
- Cardiovascular Imaging
- Medical Image Analysis
- Artificial Intelligence in Medicine
Background:
- High-resolution spiral real-time cine imaging offers detailed cardiac visualization but often requires lengthy reconstruction times.
- Accurate quantification of left ventricular ejection fraction (LVEF) is crucial for diagnosing and monitoring heart conditions.
- Current deep learning (DL) methods for image reconstruction and segmentation in cardiac MRI are areas of active development.
Purpose of the Study:
- To develop and evaluate a DL-based rapid spiral image reconstruction (DESIRE) technique.
- To create and assess a DL-based segmentation approach for quantifying LVEF.
- To enable rapid, high-quality cardiac MRI reconstruction and LVEF quantification.
Main Methods:
- A 3D U-Net architecture was employed for the DESIRE image reconstruction network.
- A 2D U-Net architecture was utilized for the DL-based segmentation network.
- The DESIRE technique was compared against low-rank plus sparse (L+S) reconstruction, and DL segmentation was compared against manual contouring for LVEF quantification.
Main Results:
- The DESIRE technique demonstrated comparable image quality to L+S reconstruction, with no statistically significant difference.
- DL-based segmentation for LVEF quantification showed no significant difference compared to manual segmentation (p > 0.05).
- DESIRE reconstruction time was approximately 32 seconds per dynamic series, significantly faster than L+S (approx. 3 minutes), and DL segmentation took less than 5 seconds.
Conclusions:
- The proposed DL-based DESIRE reconstruction and segmentation methods enable rapid, whole-heart image reconstruction within one minute.
- These techniques provide accurate and efficient quantification of left ventricular function in high-resolution spiral real-time cine MRI.
- The study highlights the potential of DL for accelerating cardiac MRI workflows and improving diagnostic efficiency.
