High-dimensional fast convolutional framework (HICU) for calibrationless MRI.

Area of Science:

• Medical Imaging
• Computational Science
• Biomedical Engineering

Background:

• Calibrationless magnetic resonance imaging (Cl-MRI) offers high acceleration and flexible sampling but faces computational challenges.
• Existing Cl-MRI methods are often limited by computational complexity and large memory requirements, hindering clinical use.

Purpose of the Study:

• To introduce a novel computational procedure for accelerated, calibrationless magnetic resonance image (Cl-MRI) reconstruction.
• The procedure aims to be fast, memory-efficient, and scalable for high-dimensional imaging.

Main Methods:

• The HIgh-dimensional fast convolutional framework (HICU) was developed for rapid and memory-efficient recovery of unsampled k-space data.
• HICU was applied to diverse datasets including 2D brain, cardiac cine, 3D knee, and multi-shot diffusion-weighted imaging (MSDWI).

Main Results:

• HICU demonstrated 1-2 orders of magnitude speedup in 2D imaging compared to other Cl-MRI methods, maintaining image quality.
• For 2D cine and 3D imaging, HICU achieved comparable computation times to SENSE-based compressed sensing, with improved signal-to-error ratio and perceptual quality.
• Feasibility of HICU was confirmed for challenging MSDWI applications.

Conclusions:

• The HICU method provides efficient computation and scalability for MRI reconstruction.
• HICU is extendable to a wide range of magnetic resonance imaging applications.