Duohua Sun¹, Jean-Philippe Galons², Chidi Ugonna¹, Silu Han¹, Mahesh Keerthivasan³, Marc Lindley⁴, and Nan-kuei Chen^1
1Biomedical Engineering, The University of Arizona, Tucson, AZ, United States, ²Medical Imaging, The University of Arizona, Tucson, AZ, United States, ³Siemens healthineers, Tucson, AZ, United States, ⁴GE Healthcare, Waukesha, WI, United States

We present an approach for improving spatial and temporal resolution of complex-valued T₂^*-weighted dynamic MRI. Compared with the conventional magnitude-valued super-resolution approaches, our technique utilizes phase information to better recover signal loss caused by susceptibility gradients and generate finer representations of temporal dynamic signal variation. Results from numerical and hybrid simulation show that promising improvements in image resolution, susceptibility artifact reduction and temporal signal variation representation can be achieved using our complex-valued super-resolution MRI scheme when compared to magnitude-valued super-resolution. 

Yilong Liu^1,2, Zheyuan Yi^1,2,3, Yujiao Zhao^1,2, Hua Guo⁴, and Ed X. Wu^1,2
1Laboratory of Biomedical Imaging and Signal Processing, The University of Hong Kong, Hong Kong, China, ²Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, China, ³Department of Electrical and Electronic Engineering, Southern University of Science and Technology, Shenzhen, China, ⁴Center for Biomedical Imaging Research, Department of Biomedical Engineering, Tsinghua University, Beijing, China

This study presents a calibrationless multi-slice spiral MRI reconstruction method based on low rank Hankel tensor completion (MS-HTC). In this study, the sampling pattern of adjacent slices complements each other by using the spiral trajectories of different rotation angles, and MS-HTC exploits the similarities of coil sensitivities, spatial support, and image content. The proposed method was evaluated with human brain spin-echo spiral MR data. The results show that MS-HTC can significantly reduce residual error compared to single-slice reconstruction with simultaneous autocalibrating and k-space estimation (SAKE).

Berkin Bilgic¹, Benedikt A Poser², Christian Langkammer³, Kawin Setsompop¹, and Congyu Liao^1
1Martinos Center for Biomedical Imaging, Charlestown, MA, United States, ²Maastricht University, Maastricht, Netherlands, ³Department of Neurology, Medical University of Graz, Graz, Austria

We introduce 3D blip-up and -down acquisition (3D-BUDA) for 3D echo planar imaging (3D-EPI). We acquire two-shots of 3D-EPI with alternating phase-encoding to estimate B₀ information. Incorporating this into the joint reconstruction of the shots eliminates distortion and enables signal averaging, permitting a 22-second, high-SNR acquisition at 1 mm³ resolution. While shifted sampling between the shots provides complementary k-space coverage, using low-rank regularization eliminates shot-to-shot variations. SNR gain of 7T allows for additional partition acceleration, enabling a 9-second whole-brain scan at R_inplane×R_z=5×2. These are combined with a self-supervised dipole inversion algorithm for Quantitative Susceptibility Mapping (QSM) which outperforms state-of-the-art reconstructions.

Li Feng¹ and Fang Liu^2,3
1Biomedical Engineering and Imaging Institute and Department of Radiology, Icahn School of Medicine at Mount Sinai, New York, NY, United States, ²Gordon Center for Medical Imaging, Massachusetts General Hospital, Harvard Medical School, Boston, MA, United States, ³Department of Radiology, University of Wisconsin Madison, Madison, WI, United States

This work presents a free-breathing motion-resolved golden-angle image reconstruction method called XD-GRASP-Pro, which extends the original XD-GRASP (eXtra-Dimensional Golden-angle RAdial Sparse Parallel MRI) method with imProved reconstruction performance through an additional self-estimated/calibrated low-rank subspace-constraint. The temporal basis used to construct the subspace is estimated from an intermediate reconstruction step on the low-resolution portion of radial k-space, which eliminates the need of using auxiliary data or a physical signal model that is not always available. XD-GRASP-Pro were tested for high spatiotemporal resolution motion-resolved liver MRI.

Niek R.F. Huttinga^1,2, Tom Bruijnen^1,2, Cornelis A.T. van den Berg^1,2, and Alessandro Sbrizzi^1,2
1Computational Imaging Group for MR diagnostics and Therapy, Center for Image Sciences, University Medical Center Utrecht, Utrecht, Netherlands, ²Department of Radiology, Division of Imaging and Oncology, University Medical Center Utrecht, Utrecht, Netherlands

We propose low-rank MR-MOTUS, a framework for real-time reconstructions of 3D respiratory motion-fields for MR-guided radiotherapy. Low-rank MR-MOTUS factorizes space-time motion-fields into static spatial components and dynamic temporal components. This allows to 1) exploit spatial and temporal correlations in motion, and 2) split the reconstruction into a large-scale off-line training phase, and a small-scale on-line inference phase.  Results show that in the on-line inference phase 3D respiratory motion can be estimated in 130ms, from data acquired in 24ms. This yields a total latency of 154ms, and low-rank MR-MOTUS thereby paves the way for real-time MR-guided radiotherapy on the MR-linac.

Christopher M. Sandino¹, Frank Ong¹, and Shreyas S. Vasanawala^2
1Department of Electrical Engineering, Stanford University, Stanford, CA, United States, ²Department of Radiology, Stanford University, Stanford, CA, United States

Unrolled neural networks (UNNs) have surpassed state-of-the-art methods for dynamic MR image reconstruction from undersampled k-space measurements. However, 3D UNNs suffer from high computational complexity and memory demands, which limit applicability to large-scale reconstruction problems. Previously, subspace learning methods have leveraged low-rank tensor models to reduce their memory footprint by  reconstructing simpler spatial and temporal basis functions. Here, a deep subspace learning reconstruction (DSLR) framework is proposed to learn iterative procedures for estimating these basis functions. As proof of concept, we train DSLR to reconstruct undersampled cardiac cine data with 5X faster reconstruction time than a standard 3D UNN.

Anne Slawig¹ and Herbert Köstler^1
1Departement of Diagnostic and Interventional Radiology, University Hospital Würzburg, Würzburg, Germany

Balanced steady state free precession sequences are well suited for cardiac imaging as they are fast, yield high signal and provide excellent contrast between blood and myocardium. To avoid the typical banding artifacts in such sequences, conventionally, multiple phase-cycled acquisitions are performed and combined to one image. Using a frequency-modulated bSSFP sequence the acquisition of many off-resonances can be performed in one single scan and a model-based reconstruction using the elliptical signal model performs well to reconstruct such measurements. Therefore, it allows the reconstruction of multiple off-resonance states for multiple heart phases from a single frequency-modulated bSSFP measurement.

Andrew Phair¹, Michael Brideson¹, Jin Jin^2,3,4, Mingyan Li², Stuart Crozier², and Lawrence Forbes^1
1School of Natural Sciences, University of Tasmania, Hobart, Australia, ²School of Information Technology and Electrical Engineering, University of Queensland, Brisbane, Australia, ³ARC Training Centre for Innovation in Biomedial Imaging Technology, University of Queensland, Brisbane, Australia, ⁴Mark and Mary Stevens Neuroimaging and Informatics Institute, University of Southern California, Los Angeles, CA, United States

We present WARF, a novel reconstruction algorithm for radial MRI data acquired with a rotating radio-frequency coil (RRFC). The algorithm reconstructs each pixel as a weighted sum of all acquired data, with the weights determined by the k-space sampling pattern. The theory behind WARF leading to the derivation of appropriate weights is presented, and then WARF is applied to both simulated and experimental data sets. The results indicate WARF is achieving an improved robustness to RRFC angular velocity variability and k-space trajectory deviation compared with existing reconstruction methods.

Kerstin Hammernik¹, Gastao Cruz², Thomas Kuestner², Claudia Prieto², and Daniel Rueckert^1
1Department of Computing, Imperial College London, London, United Kingdom, ²School of Biomedical Engineering and Imaging Sciences, Kings College London, London, United Kingdom

In this work, we study the influence of prior knowledge in learning-based non-Cartesian 2D CINE MR image reconstruction. The proposed approach uses a novel minimal deep learning setup to embed the acquired non-Cartesian multi-coil data and conventional spatio-temporal (3D and 2D+t) Fields-of-Experts regularization in a proximal gradient variational network, achieving promising results for up to 12-fold retrospectively undersampled tiny golden-angle radial CINE imaging.

Zijing Zhang^1,2, Congyu Liao², Jaejin Cho², Mary Kate Manhard², Wei-Ching Lo³, Jinmin Xu^1,2, Kawin Setsomepop^2,4,5, Huafeng Liu¹, and Berkin Bilgic^2,4
1State Key Laboratory of Modern Optical Instrumentation, College of Optical Science and Engineering, Zhejiang University, Hangzhou, China, ²Department of Radiology, A. A. Martinos Center for Biomedical Imaging, Massachusetts general hospital, Charlestown, MA, United States, ³Siemens Medical Solutions, Boston, MA, United States, ⁴Harvard Medical School, Boston, MA, United States, ⁵Harvard-MIT Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge, MA, United States

We propose dSAGE, a new EPI sequence for diffusion, Spin- and Gradient-echo imaging. We exploit unused sequence time during the b=0 acquisition in a diffusion experiment to collect additional T2*- and T2'-weighted contrasts with high in-plane resolution for free. We use a multi-shot acquisition with high in-plane acceleration to achieve 1x1 mm2 resolution, and alternate the phase-encoding polarities across the shots to eliminate geometric distortion using the navigator-free Blip Up-Down Acquisition(BUDA) technique(1). We demonstrate the ability of BUDA-dSAGE to provide whole-brain, distortion-free, high-SNR images with T2*-, T2'-, T2-weighted contrasts and 3-direction dMRI and apparent diffusion coefficient (ADC) maps in 1-minute.