Anjali Datta¹, Dwight Nishimura¹, and Corey Baron^2
1Electrical Engineering, Stanford, Stanford, CA, United States, ²Medical Biophysics, Western University, London, ON, Canada

For banding-free bSSFP cardiac cine, a highly-accelerated projection-reconstruction sequence acquires three phase-cycles within a short breathhold.  Data is also acquired on the rewinds, enabling generation of B₀ maps using BMART, which are used for phase-cycle combination.  We show that this data is well-captured by a multi-scale low-rank (MSLR) model, which recovers the normal and rewind images from the aggressively-undersampled data with less streaking and blurring than total-variation-regularized ESPIRiT.  In addition to improving the phase-cycle component images, MSLR facilitates generation of temporally-resolved B₀ maps with good SNR.  Together, these two improvements result in the final, field-map-combined cine images having high quality.

Yichen Hu¹ and Junpu Hu^2
1UIH America, Inc., Houston, TX, United States, ²United Imaging Healthcare, Shanghai, China

We applied MDI algorithm to classic PSIR reconstruction for cardiac imaging and demonstrated the feasibility and effectiveness of the approach for improved SNR in phase-sensitive contrast imaging. In comparison to the conventional reconstruction, the algorithm offers a simplified and fast pathway to achieve desired image contrast. Improved SNR in the T1W images in comparison to the conventional PSIR reconstruction was obtained.

Pei Han^1,2, Junzhou Chen^1,2, Fei Han³, Zhehao Hu^1,2, Debiao Li^1,2, Anthony G. Christodoulou^1,2, and Zhaoyang Fan^1,4
1Biomedical Imaging Research Institute, Cedars-Sinai Medical Center, Los Angeles, CA, United States, ²Department of Bioengineering, UCLA, Los Angeles, CA, United States, ³Siemens Medical Solutions USA, Inc., Los Angeles, CA, United States, ⁴Departments of Radiology and Radiation Oncology, University of Southern California, Los Angeles, CA, United States

We propose SPIDER, a new technique for real-time multi-contrast 3D imaging. A “Prep” scan is first performed to learn the static information; a “Live” scan is then performed to acquire only single k-space projection for dynamic information. With the information learned in the “Prep” scan, 3D multi-contrast images can be generated with simple matrix multiplication, which yields a latency of 50ms or less.

Zhao He¹, Ya-Nan Zhu², Suhao Qiu¹, Xiaoqun Zhang², and Yuan Feng^1
1Institute for Medical Imaging Technology, School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China, ²School of Mathematical Sciences, MOE-LSC and Institute of Natural Sciences, Shanghai Jiao Tong University, Shanghai, China

A low-rank and sparsity (LS) decomposition algorithm with framelet transform was proposed for real-time interventional MRI (i-MRI). Different from the existing LS decomposition, we exploited the spatial sparsity of both the low-rank and sparsity components. A primal dual fixed point (PDFP) method was adopted for optimization to avoid solving subproblems. We carried out intervention experiments with gelatin and brain phantoms to validate the algorithm. Reconstruction results showed that the proposed method can achieve an acceleration of 40 folds.

Runyu Yang¹, Yuze Li¹, and Huijun Chen^1
1Center for Biomedical Imaging Research, Department of Biomedical Engineering, Tsinghua University, Beijing, China

Achieving high spatio-temporal resolutions is challenging in dynamic magnetic resonance imaging (dMRI). It is effective to use low-rank prior and sparse prior jointly for dMRI reconstruction. In this study, we proposed a novel method used low rank prior which utilize a nonconvex norm and sparse prior jointly for dMRI reconstruction. The effectiveness of the proposed method was investigated in phantom and in-vivo experiments.

Mark Wright¹, Bryson Dietz¹, Jihyun Yun^1,2, Eugene Yip², B Gino Fallone^1,2, and Keith Wachowicz^1,2
1Oncology, University of Alberta, Edmonton, AB, Canada, ²Medical Physics, Cross Cancer Institute, Edmonton, AB, Canada

A real-time acceleration method using Principal Component Analysis (PCA) was developed for use on hybrid MR-radiotherapy machines. Using principal components representative of the temporal changes of k-space in combination with incoherently undersampled data from past dynamic frames, the missing data from the current undersampled frame can be filled in. This allows for real-time fully-reconstructed images. Retrospective analysis on 15 fully-sampled lung cancer patients was used to test the method. Using metrics such as NMSE, pSNR and SSIM, image quality and temporal-robustness was assessed. Dice coefficient, centroid displacement and Hausdorff distance were used to test auto-contouring capabilities for target tracking effectiveness. 

Ingmar Middelhoff¹, Matthias Schlögl², Adrián Martín Fernández³, Silvio Fanzon^4,5, Kristian Bredies^4,5, and Rudolf Stollberger^1,5
1Institute of Medical Engineering, TU Graz, Graz, Austria, ²Solgenium OG, Linz, Austria, ³Department of Information and Communications Technologies, Pompeu Fabra University, Barcelona, Spain, ⁴Institute of Mathematics and Scientific Computing, NAWI Graz, University of Graz, Graz, Austria, ⁵BioTechMed-Graz, Graz, Austria

In this study we present an approach that combines sub-sampled encoding reconstruction and simultaneous object motion computation. For that purpose Optimal Transport is used as convex regularization for motion-afflicted measurements. It reconstructs explicit pixel-wise motion fields simultaneously to the image series. Results based on simulated data show that 8-frame image series can be reconstructed in great detail from 4-fold undersampled k-space series data from a single coil. The high potential of the presented method could be shown for the reconstruction of undersampled image series. For the recovery of the motion fields, further improvements are still necessary.

Yunsong Liu¹, Kawin Setsompop², and Justin P. Haldar^1
1Signal and Image Processing Institute, University of Southern California, Los Angeles, CA, United States, ²Department of Radiology, Stanford University, Stanford, CA, United States

gSlider is an efficient technique for diffusion MRI that uses multiple RF encodings to encode high-resolution spatial information along the slice dimension.  In this work, we investigate whether smooth-phase constraints can be used to reduce the required number of RF encodings.  Although smooth-phase constraints are classically used to reduce k-space sampling (partial Fourier acquisition), we believe that their use to reduce RF encoding requirements is novel.  Theoretical and simulation results demonstrate that, if optimized RF encodings are used, phase constraints can successfully be used to reduce the number of required RF encodings in image regions where the phase is smooth.

Philip Kenneth Lee^1,2, Yuxin Hu^1,2, Catherine Judith Moran², Bruce Lewis Daniel², and Brian Andrew Hargreaves^1,2,3
1Electrical Engineering, Stanford University, Stanford, CA, United States, ²Radiology, Stanford University, Stanford, CA, United States, ³Biomedical Engineering, Stanford University, Stanford, CA, United States

In diffusion weighted imaging, multi-shot Echo Planar Imaging (EPI) is preferred over single-shot EPI due to reduced geometric distortion. Recent work has shown that low-rank reconstructions can correct ghosts from shot-to-shot phase without explicitly acquiring a phase navigator. These works have been limited to EPI sampling trajectories with uniform ky sampling. 2D Cartesian Fast Spin Echo (FSE) is a distortionless alternative to multi-shot EPI that has greater freedom in k-space traversal and reduced chemical shift artifacts. Using FSE, we demonstrate in simulation and in vivo that an intelligent choice of sampling pattern greatly enhances image quality in multi-shot diffusion imaging.

Uten Yarach^1,2, Frank Godenschweger³, Matt A Bernstein², Myung-Ho In², Itthi Chatnuntawech4⁴, Kawin Setsompop⁵, Oliver Speck³, and Joshua Trzasko^2
1Radiologic Technology Department, Associated Medical Sciences, Chiang Mai University, Chinag Mai, Thailand, ²Department of Radiology, Mayo Clinic, Rochester, MN, USA, Rochester, MN, United States, ³Otto-von-Guericke University Magdeburg, Biomedical Magnetic Resonance, Magdeburg, Germany, ⁴National Nanotechnology Center (NANOTEC), National Science and Technology Development Agency (NSTDA), Bangkok, Thailand, ⁵Department of Radiology, Stanford University, Stanford, CA, United States

Short-Axis Propeller (SAP) EPI enables short echo spacing, thereby minimizing geometric distortion, while providing low resolution in the readout direction of single blades. These multiple low-resolution blades can be combined to create a final high-resolution image. However, off-resonance effect often results in blurring after blade combination. In this work, we extend a model-based framework for reconstructing SAP-EPI to minimize off-resonance induced blurring artifacts. Moreover, locally low rank (LLR) regularization is incorporated to estimate per-blade phase calibrations. As a result, the proposed technique enables high-resolution SAP-EPI images with minimizing blurring artifact and no need for phase calibration of different multi-blade directions.

Tobias Speidel¹, Patrick Metze¹, Kilian Stumpf¹, Thomas Hüfken¹, and Volker Rasche^1
1Internal Medicine II, Ulm University Hospital, Ulm, Germany

The calculation of k-space trajectories in MRI usually involves prior knowledge of the FOV, since the desired FOV defines a minimum k-space sampling density. The reconstruction of a FOV which is larger than what is represented by the primary sampling density is equal to undersampling in k-space. Arising artefacts are strictly dependent on the underlying k-space trajectory, which leads to advantages for k-space trajectories with low-coherent aliasing properties, also for the combination with non-linear reconstruction techniques. Based on a generalised form of the "Seiffert-Spirals", this abstract describes an imaging modality that does not require prior commitment to an imaging FOV.

Kirsten Koolstra¹ and Rob Remis^2
1Division of Image Processing, Leiden University Medical Center, Leiden, Netherlands, ²Circuits and Systems, Delft University of Technology, Delft, Netherlands

Long reconstruction times of compressed sensing problems can be reduced with the help of preconditioning techniques. Efficient preconditioners are often not straightforward to design. In this work, we explore the feasibility of designing a preconditioner with a neural network. We integrate the learned preconditioner in a classical reconstruction framework, Split Bregman, and compare its performance to an optimized circulant preconditioner. Results show that it is possible for a learned preconditioner to meet and slightly improve upon the performance of existing preconditioning techniques. Optimization of the training set and the network architecture is expected to improve the performance further.

Curtis A Corum^1,2, Abdul Haseeb Ahmed², Mathews Jacob², Vincent Magnotta², and Stanley Kruger^2
1Champaign Imaging LLC, Shoreview, MN, United States, ²University of Iowa, Iowa City, IA, United States

FID sequences such as the zero echo time sequence (ZTE) posses many advantages including capturing signals from fast relaxing spins, efficient use of time and quite acoustic operation. They possess one major disadvantage, information from the start of the FID is nearly always missing or corrupted due to requirements for RF pulse time and T/R switching (ZTE) and gradient ramping (UTE).

Here we modify and apply for the first time a robust and computationally efficient missing point and phase estimation algorithm originating in the solid state NMR community for ZTE imaging sequences.

Eric A Borisch¹, Roger C Grimm¹, Soudabeh Kargar², Akira Kawashima³, Joshua D Trzasko¹, and Stephen J Riederer^1
1Radiology, Mayo Clinic, Rochester, MN, United States, ²Radiology, University of Wisconsin-Madison, Madison, WI, United States, ³Radiology, Mayo Clinic, Phoenix, AZ, United States

A previously described method for producing an image stack with enhanced through-plane resolution from an acquired set of overlapping thicker 2D slices is limited by noise enhancement when a linear reconstruction is used. To improve the resulting sharpness and noise performance of the output images, a sparsity-regularized (wavelet) full forward model based iterative reconstruction is developed. Initial results with a composite-splitting gradient descent solver provide promising noise and resolution enhancement performance. Future work includes refinements to the forward model and improving optimization convergence of the solver through momentum-based algorithms.

Gabriel Varela-Mattatall^1,2 and Ravi S Menon^1,2
1Centre for Functional and Metabolic Mapping, Robarts Research Institute, Western University, London, ON, Canada, ²Department of Medical Biophysics, Schulich School of Medicine and Dentistry, Western University, London, ON, Canada

WaveCS is the combination of corkscrew trajectories in k-space with the application of compressed sensing reconstruction. However, its accurate reconstruction depends critically on user-defined tuning. This is a tedious process that may require additional optimization if acquisition parameters are changed. Furthermore, an incorrect regularization weighting could generate noise amplification, emergence of artifacts, smoothing and loss of structural information. Here, we present a fast, non-iterative and automatic procedure that estimates the regularization weighting and which reconstruction is comparable to previous reconstructions using more tedious approaches that are considered the state-of-the-art.

Shraddha Pandey^1,2, Arthur David Snider¹, Wilfrido Moreno¹, Harshan Ravi², Ali Bilgin³, and Natarajan Raghunand^2,4
1Electrical Engineering, University of South Florida, Tampa, FL, United States, ²Cancer Physiology, Moffitt Cancer Center, Tampa, FL, United States, ³Departments of Medical Imaging, Biomedical Engineering, and Electrical & Computer Engineering, University of Arizona, Tucson, AZ, United States, ⁴Department of Oncologic Sciences, University of South Florida, Tampa, FL, United States

A joint reconstruction framework is proposed to reconstruct a series of T1-weighted, T2-weighted, and T2*-weighted images and corresponding parameter maps simultaneously from undersampled cartesian k-space data. Joint Total Variation (JTV) and model-based constraints were employed to resolve the ambiguity introduced due to undersampling. T1 and T2 maps were used to identify fluid, adipose, muscle and tumor tissue types. T2*w images reconstructed from undersampled data were analyzed to produce maps of Proton Density Fat Fraction (PDFF), Proton Density Water Fraction (PDwF), and the relaxation rates of water ($$$R^*_{2w}$$$) and fat ($$$R^*_{2f}$$$) in each tissue type [1].

Gabriel Varela-Mattatall^1,2, Corey A Baron^1,2, and Ravi S Menon^1,2
1Centre for Functional and Metabolic Mapping, Robarts Research Institute, Western University, London, ON, Canada, ²Department of Medical Biophysics, Schulich School of Medicine and Dentistry, Western University, London, ON, Canada

Most compressed sensing reconstruction procedures have relied on user-defined tuning and/or multiple comparisons to the fully sampled data to demonstrate both the feasibility of compressed sensing reconstructions as the appropriate selection of the regularization weighting. Obviously, this is a time-consuming procedure which could be avoided if we had a method that provides the regularization weighting in an automatic, non-iterative, prospective, and fast manner. Here, we present such method that could significantly accelerate research that is based on compressed sensing and improve its clinical translatability when the sparsifying domain is based on the wavelet transform. 

Qingyong Zhu¹, Jing Cheng², Zhuo-Xu Cui¹, and Dong Liang^1,2
1Research Center for Medical AI, SIAT, Chinese Academy of Sciences, Shenzhen, China, ²Paul C. Lauterbur Research Center for Biomedical Imaging, SIAT, Chinese Academy of Sciences, Shenzhen, China

The Plug-and-Play prior (PnP) is also known as denoising prior which has been successfully utilized in non-linear imaging problems. The paper presents a hybrid PnP which is incorporated into the fast composite splitting algorithm (FCSA) for compressed sensing magnetic resonance imaging (CS-MRI). The advantage of the hybrid PnP over generic PnP such as BM3D is that it can further remove artifacts and preserve adaptively fine structures in CS-MRI reconstruction. Experimental results and performance comparisons with generic PnP-FISTA show the superiority of the proposed approach even for high acceleration factor.

Nicholas Dwork¹, Corey A. Baron², Ethan M. I. Johnson³, Daniel O'Connor⁴, John M. Pauly⁵, and Peder E.Z. Larson^1
1Radiology and Biomedical Imaging, UCSF, San Francisco, CA, United States, ²Robarts Research, Western University, Ontario, ON, Canada, ³Biomedical Engineering, Northwestern University, Evanston, IL, United States, ⁴Mathematics and Statistics, University of San Francisco, San Francisco, CA, United States, ⁵Electrical Engineering, Stanford University, Stanford, CA, United States

We present a fast method for generating a variable density poisson-disc sampling pattern. A minimum parameter value is used to create a background grid array for keeping track of those points that might affect any new candidate point; this reduces the number of conflicts that must be checked before acceptance of a new point, thus reducing the number of computations required.  We demonstrate the algorithm's ability to generate variable density poisson-disc sampling patterns where the variations in density are a function of direction.  We further show that these sampling patterns are appropriate for compressed sensing applications.

Gabriel Varela-Mattatall^1,2, Corey A Baron^1,2, and Ravi S Menon^1,2
1Centre for Functional and Metabolic Mapping, Robarts Research Institute, Western University, London, ON, Canada, ²Department of Medical Biophysics, Schulich School of Medicine and Dentistry, Western University, London, ON, Canada

Low rank is an appealing method to reconstruct multiple images that share common properties between them. The highest variance, from a singular value decomposition perspective, comes from acquisition noise; therefore, noise can be tracked and discarded by selecting either the ideal rank or denoising threshold. However, the a priori determination of either of them is still an open question. In this work, we develop a general, non-iterative, fast, and automatic procedure to determine the regularization weighting for low rank reconstruction problems.