Resolution-Robust 3D MRI Reconstruction with 2D Diffusion Priors: Diverse-Resolution Training Outperforms Interpolation

TL;DR

This paper introduces a resolution-robust 3D MRI reconstruction method using 2D diffusion priors, demonstrating that training on multiple resolutions outperforms interpolation and model-based approaches in clinical scenarios.

Contribution

The paper proposes a simple, data-centric training approach for 2D diffusion models on multiple resolutions, achieving robust 3D MRI reconstruction across varying voxel sizes.

Findings

Training on multiple resolutions improves robustness to voxel size variations.

Model-based approaches like Gaussian splatting and neural representations do not close the performance gap.

The proposed diffusion-guided regularization method achieves competitive quality across resolutions.

Abstract

Deep learning-based 3D imaging, in particular magnetic resonance imaging (MRI), is challenging because of limited availability of 3D training data. Therefore, 2D diffusion models trained on 2D slices are starting to be leveraged for 3D MRI reconstruction. However, as we show in this paper, existing methods pertain to a fixed voxel size, and performance degrades when the voxel size is varied, as it is often the case in clinical practice. In this paper, we propose and study several approaches for resolution-robust 3D MRI reconstruction with 2D diffusion priors. As a result of this investigation, we obtain a simple resolution-robust variational 3D reconstruction approach based on diffusion-guided regularization of randomly sampled 2D slices. This method provides competitive reconstruction quality compared to posterior sampling baselines. Towards resolving the sensitivity to resolution-shifts, we investigate state-of-the-art model-based approaches including Gaussian splatting, neural representations, and infinite-dimensional diffusion models, as well as a simple data-centric approach of training the diffusion model on several resolutions. Our experiments demonstrate that the model-based approaches fail to close the performance gap in 3D MRI. In contrast, the data-centric approach of training the diffusion model on various resolutions effectively provides a resolution-robust method without compromising accuracy.