Low-rank plus sparse matrix decomposition for accelerated dynamic MRI with separation of background and dynamic...

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Signal Processing

Background:

• Dynamic MRI data is often undersampled, limiting spatial and temporal resolution.
• Conventional reconstruction methods struggle with motion-sensitive background subtraction.

Purpose of the Study:

• To apply the low-rank plus sparse (L+S) matrix decomposition model for dynamic MRI reconstruction.
• To reconstruct undersampled dynamic MRI as a superposition of background and dynamic components.
• To evaluate L+S model performance in various clinical applications.

Main Methods:

• Utilized a convex optimization approach for multicoil L+S reconstruction.
• Employed nuclear norm for low-rank enforcement (L) and l1 norm for sparsity enforcement (S).
• Tested feasibility in cardiac perfusion, cine, time-resolved angiography, and abdominal/breast perfusion MRI with Cartesian and radial sampling.

Main Results:

• The L+S model significantly increased dynamic MRI data compressibility, allowing for high acceleration factors.
• Achieved superior background suppression compared to conventional data subtraction, which is prone to motion artifacts.
• Demonstrated effective reconstruction for various dynamic MRI applications.

Conclusions:

• L+S model enables high acceleration, enhancing spatial and temporal resolution in dynamic MRI.
• Provides effective background suppression without requiring subtraction or complex modeling.
• Promises to improve the clinical utility of dynamic MRI techniques.