Group-Patch Joint Compression: Compressing Dynamic B₀ and Static RF Spatial Modulations Across k-Space Subregion...

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Medical Physics
• Signal Processing

Background:

• Accelerating MRI acquisition is crucial for clinical applications.
• Rapid B0 field modulations offer potential for enhanced MRI but introduce reconstruction complexities.
• Existing methods struggle with non-Fourier encoded data, especially in compressed-sensing and neural network frameworks.

Purpose of the Study:

• To develop a novel reconstruction method for MRI with rapid B0 field modulations.
• To address the computational challenges posed by non-Fourier encoding in advanced MRI techniques.
• To improve reconstruction speed and reduce memory requirements for accelerated MRI.

Main Methods:

• A group-patch joint compression strategy was developed to exploit encoding redundancy.
• k-space data is compressed patch-by-patch, jointly encoded by B0 and RF spatial functions.
• An adapted RF array compression algorithm was used, mimicking expanded virtual receiver channels, evaluated on human brain scans at 9.4T/3T.

Main Results:

• The proposed joint compression achieved substantially higher compression factors (11x-20x) than conventional RF-only compression.
• Negligible encoding loss was observed with high compression factors, significantly reducing reconstruction time and peak memory usage.
• Compressed-sensing reconstruction times were dramatically reduced to 1.4-5.1s/slice and 177s-10.1min/volume.

Conclusions:

• Compressing multidimensional k-space patches in groups, considering joint B0 and RF encoding, outperforms RF-only compression.
• This approach effectively mitigates computational bottlenecks associated with combining rapid B0 and RF-receiver modulations.
• The method offers a significant advancement for accelerating MRI acquisition and reconstruction.