Problem
Research questions and friction points this paper is trying to address.
MRI reconstruction
non-Cartesian
accelerated imaging
distribution shift
reconstruction robustness
Weakly Convex Ridge Regularization for 3D Non-Cartesian MRI Reconstruction
🤖 AI Summary
This work addresses the challenges of high computational latency, instability of deep learning approaches, and insufficient robustness to distribution shifts in highly accelerated 3D non-Cartesian MRI reconstruction. To this end, we propose a variational reconstruction method based on rotation-invariant weakly convex ridge regularization (WCRR). Our approach introduces WCRR into MRI reconstruction for the first time, effectively integrating physical modeling with data-driven priors while maintaining theoretical guarantees and expressive power—without relying on iterative networks. Experimental results demonstrate that the proposed method consistently outperforms conventional baselines on both simulated and real prospective datasets, achieving reconstruction quality comparable to state-of-the-art DRUNet-based Plug-and-Play methods, yet with significantly improved computational efficiency and enhanced robustness to variations in acquisition protocols.
📝 Abstract
While highly accelerated non-Cartesian acquisition protocols significantly reduce scan time, they often entail long reconstruction delays. Deep learning based reconstruction methods can alleviate this, but often lack stability and robustness to distribution shifts. As an alternative, we train a rotation invariant weakly convex ridge regularizer (WCRR). The resulting variational reconstruction approach is benchmarked against state of the art methods on retrospectively simulated data and (out of distribution) on prospective GoLF SPARKLING and CAIPIRINHA acquisitions. Our approach consistently outperforms widely used baselines and achieves performance comparable to Plug and Play reconstruction with a state of the art 3D DRUNet denoiser, while offering substantially improved computational efficiency and robustness to acquisition changes. In summary, WCRR unifies the strengths of principled variational methods and modern deep learning based approaches.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Weakly Convex Regularization
Non-Cartesian MRI
Variational Reconstruction
Rotation Invariance
Deep Learning Acceleration
🔎 Similar Papers
No similar papers found.
