Accelerated partial separable model using dimension-reduced optimization technique for ultra-fast cardiac MRI

Area of Science:

• Medical Imaging
• Magnetic Resonance Imaging (MRI)
• Computational Imaging

Background:

• Dynamic object imaging with high temporal resolution is a challenge in MRI.
• The partial separable (PS) model improves image quality but requires long acquisition and reconstruction times.
• Existing PS models need acceleration for clinical viability.

Purpose of the Study:

• To accelerate the PS model for faster acquisition and reconstruction.
• To maintain high image quality during accelerated dynamic MRI.
• To enable real-time MRI applications.

Main Methods:

• Developed a dimension-reduced optimization technique for the PS model.
• Implemented optimization within a subspace to exploit dimension-reduction properties.
• Optimized the data consistency term and used Tikhonov regularization for temporal differences.
• Validated the method in free-running cardiac MRI using retrospective and prospective data.

Main Results:

• The proposed method achieved superior image quality compared to competing algorithms.
• Demonstrated robust performance with shortened acquisition times and suboptimal settings.
• Achieved a 20-fold speedup over the PS+sparse method, with reconstruction in seconds.

Conclusions:

• The accelerated PS model significantly reduces MRI scan and reconstruction times.
• This method offers a promising solution for efficient dynamic MRI.
• Potential for widespread clinical adoption in dynamic MRI and real-time applications.