Accelerated image reconstruction with separable Hankel regularization in parallel MRI

Magnetic resonance imaging has been widely adopted in clinical diagnose, however, it suffers from relatively long data acquisition time. Sparse sampling with reconstruction can speed up the data acquisition duration. As the state-of-the-art magnetic resonance imaging methods, the structured low rank reconstruction approaches embrace the advantage of holding low reconstruction errors and permit flexible undersampling patterns. However, this type of method demands intensive computations and high memory consumptions, thereby resulting in a lengthy reconstruction time. In this work, we proposed a separable Hankel low rank reconstruction method to explore the low rankness of each row and each column. Furthermore, we utilized the self-consistence and conjugate symmetry property of k-space data. The experimental results demonstrated that the proposed method outperforms the state-of-the-art approaches in terms of lower reconstruction errors and better detail preservation. Besides, the proposed method requires much less computation and memory consumption.Clinical Relevance— Parallel imaging, image reconstruction, Hankel low-rank

I. Introduction

As a non-invasive and non-radioactive imaging technique, magnetic resonance imaging (MRI) has served as an indispensable and widely used tool in the clinic [1]. However, the relatively slow acquisition speed limits its application. One of the effective ways to alleviate the prolonged data acquisition time is sparse sampling which acquires only a small subset of the k-space data.