M-Gaussian: An Magnetic Gaussian Framework for Efficient Multi-Stack MRI Reconstruction

TL;DR

M-Gaussian introduces a novel MRI reconstruction framework using 3D Gaussian Splatting that balances high-quality volumetric imaging with computational efficiency, especially for multi-stack MRI data.

Contribution

It adapts 3D Gaussian Splatting to MRI reconstruction, incorporating physics-based primitives, residual neural refinement, and multi-resolution training for improved speed and quality.

Findings

Achieves 40.31 dB PSNR on FeTA dataset

14 times faster than existing methods

First successful adaptation of 3D Gaussian Splatting to MRI

Abstract

Magnetic Resonance Imaging (MRI) is a crucial non-invasive imaging modality. In routine clinical practice, multi-stack thick-slice acquisitions are widely used to reduce scan time and motion sensitivity, particularly in challenging scenarios such as fetal brain imaging. However, the resulting severe through-plane anisotropy compromises volumetric analysis and downstream quantitative assessment, necessitating robust reconstruction of isotropic high-resolution volumes. Implicit neural representation methods, while achieving high quality, suffer from computational inefficiency due to complex network structures. We present M-Gaussian, adapting 3D Gaussian Splatting to MRI reconstruction. Our contributions include: (1) Magnetic Gaussian primitives with physics-consistent volumetric rendering, (2) neural residual field for high-frequency detail refinement, and (3) multi-resolution progressive training. Our method achieves an optimal balance between quality and speed. On the FeTA dataset, M-Gaussian achieves 40.31 dB PSNR while being 14 times faster, representing the first successful adaptation of 3D Gaussian Splatting to multi-stack MRI reconstruction.