Scale-Equivariant Unrolled Neural Networks for Data-Efficient Accelerated MRI Reconstruction

TL;DR

This paper introduces scale-equivariant unrolled neural networks for MRI reconstruction, enhancing data efficiency and robustness to scale variations, reducing reliance on fully-sampled training data, and maintaining fast inference.

Contribution

It proposes modeling proximal operators with scale-equivariant CNNs to improve data efficiency and robustness in MRI reconstruction, addressing variability in patient anatomy and scanner settings.

Findings

Outperforms state-of-the-art unrolled networks on scaled images

Maintains fast inference with improved robustness

Reduces dependence on fully-sampled training data

Abstract

Unrolled neural networks have enabled state-of-the-art reconstruction performance and fast inference times for the accelerated magnetic resonance imaging (MRI) reconstruction task. However, these approaches depend on fully-sampled scans as ground truth data which is either costly or not possible to acquire in many clinical medical imaging applications; hence, reducing dependence on data is desirable. In this work, we propose modeling the proximal operators of unrolled neural networks with scale-equivariant convolutional neural networks in order to improve the data-efficiency and robustness to drifts in scale of the images that might stem from the variability of patient anatomies or change in field-of-view across different MRI scanners. Our approach demonstrates strong improvements over the state-of-the-art unrolled neural networks under the same memory constraints both with and without data augmentations on both in-distribution and out-of-distribution scaled images without significantly increasing the train or inference time.