ISMRM Honors 2018

RSL Celebrates Award Winning Research

Summa Cum Laude

Multi-delay arterial spin labeling (ASL) more accurately detects hypoperfusion in Moyamoya disease: comparison with a normative PET/MRI database

Audrey Fan, Mohammad Khalighi, Jia Guo, Yosuke Ishii, Mirwais Wardak, Jun-Hyung Park, Bin Shen, Dawn Holley, Harsh Gandhi, Prachi Singh, Tom Haywood, Gary Steinberg, Frederick Chin, Greg Zaharchuk

We directly compared multi-delay arterial spin labeling (ASL) and standard ASL measurements of cerebral blood flow (CBF) to simultaneously acquired [15O]-PET scans on hybrid PET/MRI in Moyamoya disease. For these Moyamoya patients (N=15) with extremely long arterial transit times, multi-delay ASL outperforms standard ASL in regional correlation and reduces bias relative to PET. We also constructed a voxelwise, normative CBF database based on healthy controls (N=15) with PET/MRI, and identified regions of hypoperfusion in frontal and parietal regions of patients. Multi-delay ASL is more specific to areas of Moyamoya hypoperfusion (more similar to PET), whereas standard ASL overestimates these areas due to low signal.


Accurate and Efficient QSM Reconstruction using Deep Learning Accurate and EfficientReconstruction using Deep Learning

Enhao Gong, Berkin Bilgic, Kawin Setsompop, Audrey Fan, Greg Zaharchuk, John Pauly

Quantitative Susceptibility Mapping (QSM) is a powerful MRI technique to quantify susceptibility changes and reveal pathology such as multiple sclerosis (MS) lesions and demyelination. QSM reconstruction is very challenging because it requires solving an ill-posed deconvolution and removing the effects of a dipole kernel on tissue phases to obtain susceptibility. To address the limitations of existing QSM reconstruction methods in accuracy, stability and efficiency, an iteration-free data-driven QSM reconstruction is proposed that trains a deep learning model to approximate COSMOS QSM quantification from acquired signals and pre-processed phases. Cross-validated on in-vivo datasets with 15 single direction QSM scans and 3 COSMOS QSM results from 3 healthy subjects, the proposed deep learning method achieves accurate QSM reconstruction, outperforming state-of-the-art methods across various metrics. The deep learning solution is also faster than iterative reconstruction by several orders of magnitude, which enables broader clinical applications.


Accelerated T2-Weighted Imaging of the Abdomen with Self-Calibrating Wave-Encoded 3D Fast Spin Echo Sequences

Feiyu Chen, Valentina Taviani, Joseph Cheng, John Pauly, Shreyas Vasanawala

In this work, a self-calibrating wave-encoded 3D FSE technique was proposed with self-refocusing gradient waveforms and autocalibrated estimation of wave-encoding point-spread-function and coil sensitivity maps. Compared to conventional Cartesian approach at the same acceleration factor, the proposed method achieves reduced artifacts and better anatomical delineation for highly undersampled abdominal imaging. It enables 10-fold acceleration for 3D FSE scans of the abdomen, allowing 3D FSE sequences to be less sensitive to subject motion.


Phase Encoded xSPEN: A High-Definition Approach to Volumetric MRI with Unusually High Acceleration Factors

Zhiyong Zhang, Michael Lustig, Lucio Frydman

xSPEN is a single-shot MRI approach whose timing and pre-acquisition hyperbolic phase, endow it with exceptional resilience to offsets. We recently introduced multi-scan, phase-encoded (PE) 3D xSPEN MRI which preserves this while increasing resolution along the PE (y) and slab-selection (z) dimensions. It is here shown that parallel receivers endow this 3D approach with unprecedented PE downsampling performances. This reflects xSPEN’s unusual kernel, whose hyperbolic phase couples the directly-sampled kzinformation with the y coil sensors. This mitigates the artifacts associated with a highly undersampled ky axis, as demonstrated by highly accelerated in vitro and human scans.


Improving Variable-Density Single-Shot Fast Spin Echo with Deep-Learning Reconstruction Using Variational Networks

Feiyu Chen, Valentina Taviani, Itzik Malkiel, Joseph Cheng, Jamil Shaikh, Stephanie Chang, Christopher Hardy, John Pauly, Shreyas Vasanawala

In this work, a deep-learning-based reconstruction approach using a variational network (VN) was developed to accelerate the variable density single-shot fast spin echo (VD SSFSE) reconstruction. The image quality of this approach was clinically evaluated compared to standard parallel imaging and compressed sensing (PICS). The VN approach achieves improved image quality with higher perceived signal-to-noise ratio and sharpness. It also allows real-time image reconstruction of VD SSFSE sequences for practical clinical deployment.

Improved Speed and Image Quality for Imaging of Liver Lesions with Auto-calibrated Wave Encoded Variable Density Single-Shot Fast Spin Echo.

Jamil Shaikh, Feiyu Chen, Valentina Taviani, Kim Vu, Shreyas Vasanawala

Abdominal T2-weighted imaging is conventionally lengthy, but single shot approaches significantly improve current acquisition times. For single shot fast spin echo (SSFSE), axial imaging speed and sharpness are constrained by limited parallel imaging acceleration.  Here, SSFSE technique with wave encoding and variable-density sampling (wSSFSE) was developed to enable higher accelerations and improve overall image quality. The purpose of this study is to assess image quality, delineation of anatomical structures, lesion conspicuity, and speed improvements with wSSFSE. 


Magna Cum Laude

Super-Resolution Musculoskeletal MRI using Deep Learning

Akshay Chaudhari, Zhognan Fang, Feliks Kogan, Jeff Wood, Kathryn Stevens, Jin Hyung Lee, Garry Gold, Brian Hargreaves

Near-isotropic high-resolution magnetic resonance imaging (MRI) of the knee is beneficial for reducing partial volume effects and allowing multi-planar image analysis. However, previous methods exploring isotropic resolutions, typically compromised in-plane resolution for thin slices, due to intrinsic signal-to-noise ratio (SNR) limitations. Even computer-vision-based super-resolution methods have been rarely been used in medical imaging due to limited resolution improvements. In this study, we utilize deep-learning-based 3D super-resolution for rapidly generating high-resolution thin-slice knee MRI from slices originally 2-8 times thicker. Through quantitative image quality metrics and a reader study, we demonstrate superior performance to both conventionally utilized and state-of-the-art super-resolution methods. 


Accelerated abdominal 4D flow MRI using 3D golden-angle cones trajectory

Christopher Sandino, Joseph Cheng, Marcus Alley, Michael Carl, Shreyas Vasanawala

4D flow MRI enables comprehensive abdominal evaluation, but long acquisition times and motion corruption limit its clinical applicability. To address these limitations, we present a 4D flow sequence with a 3D golden-angle reordered cones sampling trajectory. Cones has high sampling efficiency to allow for vastly accelerated scan times, and excellent aliasing properties that diffuse respiratory and bowel motion artifacts. To further improve motion-robustness, respiratory signals are estimated from each cone readout, and then used to suppress motion during reconstruction. We show that these techniques can be combined to achieve high quality abdominal 4D flow renderings in under 5 minutes.


Nonrigid Motion Correction using 3D iNAVs with Generalized Motion Compensated Reconstruction and Autofocusing

Srivathsan Koundinyan, Corey Baron, Nicholas Dwork, Joseph Cheng, Dwight Nishimura

We present a novel framework to combine two well-known methods for motion correction: generalized motion compensated reconstruction (GMCR) and autofocusing. In this hybrid technique, 3D image-based navigators (3D iNAVs) are utilized for motion tracking. The beat-to-beat and voxel-by-voxel motion information within the 3D iNAVs are directly inputted into GMCR to mitigate motion artifacts. To reduce computation time, an autofocusing step is incorporated. The overall correction scheme is evaluated in free-breathing coronary magnetic resonance angiography and renal magnetic resonance angiography exams. In all six in vivo studies, images reconstructed with the proposed strategy outperform those generated with beat-to-beat 3D translational correction.


Deep Learning Method for Non-Cartesian Off-resonance Artifact Correction

David Zeng, Jamil Shaikh, Dwight Nishimura, Shreyas Vasanawala, Joseph Cheng

3D cones trajectories have the flexibility to be more scan-time efficient than 3D Cartesian trajectories, especially with long readouts. However, long readouts are subject to blurring from off-resonance, limiting the efficiency. We propose a convolutional residual network to correct for off-resonance artifacts to allow for reduced scan time. Fifteen exams were acquired with both conservative readout durations and readouts 2.4x as long. Long-readout images were corrected with the proposed method. The corrected long-readout images had non-inferior (p<0.01) reader scores in all features examined compared to conservative readout images.


A Novel Method for Fast and Efficient Measurement of Diffusion Tensor Size and Shape Distributions

Grant Yang, Jennifer McNab

We demonstrate through simulations and empirical data that it is possible to simultaneously estimate the variance of the voxel-wise diffusion tensor shape and size distributions using efficient isotropic and linear diffusion encodings on a whole-body clinical MRI scanner with whole-brain coverage at 3mm isotropic resolution in under 2 minutes. 


Simultaneous Bilateral Knee MR Imaging

Feliks Kogan, Evan Levine, Akshay Chaudhari, Uchechukwuka Monu, Kevin Epperson, Edwin Oei, Garry Gold, Brian Hargreaves

Osteoarthritis (OA) is commonly a bilateral disease. While long scan time and costs have precluded separate scanning of both knees in clinical MRI, there is evidence that bilateral examinations are beneficial for evaluation of OA changes, especially for longitudinal studies. In this study, we demonstrate that a bilateral coil-array setup can image both knees simultaneously in similar scan times as conventional unilateral knee scans with comparable image quality and quantitative accuracy. This has the potential to improve the value of MRI knee evaluations.


Perceptual Accuracy of a Mixed-Reality System for MR-Guided Breast Surgical Planning in the Operating Room

Stephanie Perkins, Michael Lin, Subashini Srinivasan, Amanda Wheeler, Brian Hargreaves, Bruce Daniel

One quarter of women who undergo lumpectomy to treat early-stage breast cancer in the United States undergo repeat surgery due to concerns that residual tumor was left behind. We have developed a supine breast MRI protocol and a system that projects a 3D “hologram” of the MR data onto a patient using the Microsoft HoloLens. The goal is to reduce the number of repeated surgeries by improving surgeons’ ability to determine tumor extent. We are conducting a pilot study in patients with palpable tumors that tests a surgeon’s ability to accurately identify tumor location via mixed-reality visualization during surgical planning.

First clinical pilot study using screen-printed flexible MRI receive coils for pediatric applications

Simone Angela Winkler, Joseph Corea, Balthazar Lechene, Kendall O'Brien, John Bonanni, Fraser Robb, Greig Scott, John Pauly, Michael Lustig, Ana Arias, Shreyas Vasanawala

Pediatric MRI is often performed suboptimally by the use of heavy, large, and relatively inflexible coil arrays that are designed and built for adult MR imaging. For the child, these arrays can be intimidating and uncomfortable, restricting breathing.  For parents, they contribute to the stress of the exam.  For pediatric caregivers for smaller children, the coils complicate placing medical support equipment. Here, we assess the use of screen printed flexible coil arrays for pediatric applications, focusing on clinical image quality and caregiver acceptance. We conclude that a flexible screen-printed MRI receive coil is likely to yield diagnostic image quality and be preferred to a traditional coil by patients, parents, and caregivers.