PS-Net: Learned Partially Separable Model for Dynamic MR Imaging

TL;DR

This paper introduces a learned low-rank model for dynamic MR imaging that adaptively captures low-rank priors using a neural network, outperforming existing methods in accuracy and quality.

Contribution

It unrolls an HQS algorithm for the PS model into a learnable network with a null-space transform, enabling adaptive low-rank characterization.

Findings

Outperforms state-of-the-art CS and deep learning methods

Demonstrates superior quantitative and qualitative results on cardiac cine data

Shows effectiveness of learned low-rank regularization in dynamic MR imaging

Abstract

Deep learning methods driven by the low-rank regularization have achieved attractive performance in dynamic magnetic resonance (MR) imaging. However, most of these methods represent low-rank prior by hand-crafted nuclear norm, which cannot accurately approximate the low-rank prior over the entire dataset through a fixed regularization parameter. In this paper, we propose a learned low-rank method for dynamic MR imaging. In particular, we unrolled the semi-quadratic splitting method (HQS) algorithm for the partially separable (PS) model to a network, in which the low-rank is adaptively characterized by a learnable null-space transform. Experiments on the cardiac cine dataset show that the proposed model outperforms the state-of-the-art compressed sensing (CS) methods and existing deep learning methods both quantitatively and qualitatively.