Group-patch joint compression for highly accelerated MRI: compressing dynamic B0 and static RF spatial modulations across k-space subregion groups

TL;DR

This paper introduces a novel group-patch joint compression method for accelerated MRI that efficiently compresses dynamic B0 and static RF modulations across k-space subregions, significantly reducing reconstruction time and memory usage.

Contribution

The paper presents a new compression approach that jointly encodes B0 and RF spatial modulations across k-space patches, outperforming traditional RF-only compression methods.

Findings

Achieves 11x-20x compression factors with minimal encoding loss.

Reduces reconstruction time from minutes to seconds.

Dramatically decreases peak memory usage during MRI reconstruction.

Abstract

Purpose: To accelerate MRI further, rapid B0 field modulations can be applied during oversampled readout to capture additional physical information, e.g., Wave-CAIPI, FRONSAC, local B0 coils modulations. These methods, however, introduce additional non-Fourier-encoded dimension that cannot be resolved by FFT, posing significant reconstruction challenges particularly in compressed-sensing or neural-network frameworks. Theory and methods: Because B0 modulations vary slowly relative to the oversampled ADC dwell time, we exploit this encoding redundancy by compressing k-space patch-by-patch across subregions, each of which is jointly encoded by a distinct subset of B0 and RF (receive) spatial encoding functions. For each patch, a compression matrix is computed once and reused to compress all patches encoded by the same B0/RF spatial modulations. This can be implemented by feeding subsets of B0 and RF spatial encoding maps into an adapted conventional RF array compression algorithm, mimicking an expanded set of virtual receiver channels. This approach was evaluated on ex-vivo/in-vivo human brain scans at 9.4T/3T. Results: The proposed group-patch joint compression achieves substantially higher compression factors than conventional RF-only compression, while minimally compromising encoding efficiency. Typically, joint compression factors of 11x-20x led to negligible encoding loss, dramatically reducing reconstruction time and peak memory usage. For example, compressed-sensing reconstruction took 1.4-5.1s/2D slice, 177s-10.1min/3D volume, on a high-memory CPU node. Conclusion: Given joint encoding of dynamic B0 and static RF fields, compressing multidimensional k-space patches in separate groups outperforms compressing RF receiver channels alone. This substantially mitigates a fundamental computational bottleneck that arises when combining rapid B0 and RF-receiver modulations.