Search papers, labs, and topics across Lattice.
This paper introduces a zero-shot MRI super-resolution (SR) framework that balances data requirements and computational efficiency by using explicit 3D Gaussian representation. The method incorporates MRI-tailored Gaussian parameters to embed tissue physical properties, reducing the number of learnable parameters and preserving MR signal fidelity. A physics-grounded volume rendering strategy models MRI signal formation, and a brick-based order-independent rasterization scheme enables parallel 3D computation, leading to improved reconstruction quality and efficiency on public MRI datasets.
Achieve state-of-the-art MRI super-resolution without paired training data by encoding physical properties into 3D Gaussians and rendering them efficiently.
High-resolution Magnetic Resonance Imaging (MRI) is vital for clinical diagnosis but limited by long acquisition times and motion artifacts. Super-resolution (SR) reconstructs low-resolution scans into high-resolution images, yet existing methods are mutually constrained: paired-data methods achieve efficiency only by relying on costly aligned datasets, while implicit neural representation approaches avoid such data needs at the expense of heavy computation. We propose a zero-shot MRI SR framework using explicit Gaussian representation to balance data requirements and efficiency. MRI-tailored Gaussian parameters embed tissue physical properties, reducing learnable parameters while preserving MR signal fidelity. A physics-grounded volume rendering strategy models MRI signal formation via normalized Gaussian aggregation. Additionally, a brick-based order-independent rasterization scheme enables highly parallel 3D computation, lowering training and inference costs. Experiments on two public MRI datasets show superior reconstruction quality and efficiency, demonstrating the method's potential for clinical MRI SR.
