Reproduce “Non-rigid Motion Correction for MRI Reconstruction via Coarse-To-Fine Diffusion Models” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Frederic Wang, Jonathan I. Tamir
A new method from Caltech and UT Austin integrates a coarse-to-fine diffusion model within an alternating minimization framework for MRI reconstruction. This approach jointly corrects non-rigid motion artifacts and reconstructs images, achieving superior visual quality and lower NRMSEs on simulated data, alongside improved detail recovery in real cardiac MRI compared to existing baselines.
Key insights
- Denoising lower frequency k-space components earlier in the diffusion process provides a more stable, global image estimate, offering a better inductive bias for robust initial motion estimation.
- An alternating minimization scheme effectively decouples and jointly optimizes the complex problems of image reconstruction and non-rigid motion field estimation, enhancing overall performance.
- The C2F diffusion model's frequency-weighted noise corruption, combined with a dedicated, unbiased motion estimation step, improves convergence and accuracy for complex, undersampled motion correction.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.