View Presentation Video

Xiao-Hong Zhu¹, Rong Guo^2,3, Hannes M Wiesner¹, Yudu Li^3,4, Yibo Zhao^3,5, Wen Jin^3,5, Zhi-Pei Liang^3,5, and Wei Chen^1
1Center for Magnetic Resonance Research, Department of Radiology, University of Minnesota, Minneapolis, MN, United States, ²Siemens Medical Solutions USA, Inc., Urbana, IL, United States, ³Beckman Institute for Advanced Science and Technology, University of Illinois at Urbana-Champaign, Urbana, IL, United States, ⁴National Center for Supercomputing Applications, University of Illinois at Urbana-Champaign, Urbana, IL, United States, ⁵Department of Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, Urbana, IL, United States

Keywords: Quantitative Imaging, Metabolism, Human Brain NAD Imaging

Nicotinamide adenine dinucleotide (NAD) is an important molecule critical for cellular metabolism and signaling. The cerebral NAD concentration is only at submillimolar level, and thus, difficult to measure and imaging in vivo. In this study, we utilize 3D ³¹P MRSI-based NAD assay in combination with a SPICE-based signal processing method to evaluate the feasibility for non-invasively mapping the NAD contents in healthy human brains at 7T. The results suggest that by taking advantage of ultrahigh field scanner and advanced post-processing, it is possible to obtain whole-brain NAD maps with NAD measurements consistent with the literature reported values.

View Presentation Video

Rachel Wagner^1,2, Michael T Jurkiewicz^2,3, Jennifer Chen^4,5, Angela Jerath^6,7, Marat Slessarev^2,5,8, and Udunna Anazodo^2,9
1Neuroscience, Western University, London, ON, Canada, ²Lawson Health Research Institute, London, ON, Canada, ³Medical Imaging, London Health Sciences Centre, London, ON, Canada, ⁴Critical Care Medicine, Western University, London, ON, Canada, ⁵Western Institute for Neuroscience, Western University, London, ON, Canada, ⁶Sunnybrook Health Sciences Centre, Toronto, ON, Canada, ⁷Evaluative Clinical Sciences, Schulich Heart Program, Sunnybrook Research Institute, Toronto, ON, Canada, ⁸Critical Care Medicine, Schulich School of Medicine & Dentistry, Western University, London, ON, Canada, ⁹Department of Neurology and Neurosurgery, McGill University, Montreal, QC, Canada

Keywords: Infectious disease, COVID-19, Neuroscience

While most COVID-19 patients present with mild symptoms, many of those hospitalized with severe infection develop acute respiratory distress syndrome (ARDS)¹ and frequently report central nervous system dysfunction².  While previous research has characterized clinical symptomatology, investigation regarding brain changes remains limited. To address this gap, this review summarized brain injuries from 372 COVID-19-ARDS and 370 ARDS patients and pooled findings. COVID-19-ARDS patients presented an unusual pattern of injury preferentially targeting white matter, suggesting impact to brain connectivity. The long-term consequences of white matter injury in COVID-19-ARDS should be taken into account in devising policies for managing patient health.

View Presentation Video

Anna Lavrova¹, Jacob Richardson¹, Ryo Kurokawa¹, Mariko Kurokawa¹, Pedro Itriago-Leon², Vikas Gulani¹, Hero Hussain¹, Katherine Wright¹, Toshio Moritani¹, and Nicole Seiberlich^1
1Department of Radiology, University of Michigan, Ann Arbor, MI, United States, ²Siemens Medical Solutions USA Inc., Houston, TX, United States

Keywords: Low-Field MRI, Brain

The purpose of this study is to assess the quality of brain imaging studies performed on an FDA-approved commercial 0.55T MRI system, and to provide information about the feasibility of using this scanner in a clinical workflow. The image quality of 1356 image series (378 in healthy subjects, 978 in patients) was independently rated by two neuroradiologists. While images from T1w SPACE, DWI/ADC, FLAIR, SWI, and T2w TSE sequences received acceptable image quality scores, contrast-enhanced T1w TSE sequence received lower, although acceptable, average scores. Our results suggest that neurological MRI studies for some indications can be performed  at 0.55T.

View Presentation Video

Dushyant Kumar¹, Blake Benyard¹, Narayan Datt Soni¹, Anshuman Swain¹, Neil Wilson¹, and Ravinder Reddy^1
1Radiology, University of Pennsylvania, Philadelphia, PA, United States

Keywords: Data Acquisition, Brain, Lipid imaging

Brain lipid imaging using steady-state nuclear Overhauser effect (ssNOE), though a traditionally popular approach, suffers from multiple confounding non-NOE specific sources, including direct saturation, magnetization transfer (MT), and relevant chemical exchange species. B₀, B₁⁺- dependent data are also needed to correct for the effect of (B₀, B₁⁺)- inhomogeneities¸ leading to other issues such as patient tolerability due to substantially increased scan time. Here, we demonstrate the feasibility of brain lipid mapping using an easily implementable transient NOE (tNOE) approach for the first time. Advantages include improved specificity, faster scan time and robust quantification with minimal confounding contributions.

View Presentation Video

Suhyung Park¹, Alexander Beckett^2,3, Suvi Hakkinen³, Samantha Ma⁴, and David Feinberg^2,3
1Department of Computer Engineering, Chonnam National University, Gwangju, Korea, Republic of, ²Advanced MRI Technologies, Sebastopol, CA, United States, ³University of California, Berkeley, Berkeley, CA, United States, ⁴Siemens Medical Solutions USA, Inc, Berkeley, CA, United States

Keywords: Data Acquisition, Brain

There are significant benefits to segmented 3D EPI fMRI acquisitions, which acquires high spatial and temporal resolution across the whole brain to better understand brain neuronal activity. This can be achieved through accelerated 3D EPI imaging, which provides rapid whole-brain coverage at the cost of SNR efficiency resulting from frame-by-frame reconstruction. Here, we utilize temporal information for whole-brain coverage on 8-fold accelerated 7T fMRI acquisition without altering the subsequent fMRI results.

View Presentation Video

Wei-Tang Chang^1,2,3, Stephanie Langella⁴, Min Sung Seo⁵, Weili Lin², and Kelly Giovanello^5
1Biomedical Engineering, UNC at Chapel Hill, Chapel Hill, NC, United States, ²Biomedical Research Imaging Center, UNC at Chapel Hill, Chapel Hill, NC, United States, ³Radiology, UNC at Chapel Hill, Chapel Hill, NC, United States, ⁴Psychiatry, Harvard Medical School, Boston, MA, United States, ⁵Psychology & Neuroscience, UNC at Chapel Hill, Chapel Hill, NC, United States

Keywords: Data Acquisition, Neuro

fMRI with ultrahigh resolution has typically been acquired at 7T in order to overcome the inherently low SNR at 3T. However, 7T scanners are inaccessible to the majority of investigators. In this study, we developed new acquisition and reconstruction approaches and employed the denoising method NORDIC to improve the resolution of 3T fMRI to 1 mm isotropic with whole-brain coverage and volumetric scan time of 2 seconds. Our results showed improved tSNR and low spatial blurring. The detection capability of resting-state networks was comparable to that of 7T. BOLD activations in cortical and subcortical regions were observed in task-based fMRI.

View Presentation Video

Florian Birk^1,2, Klaus Scheffler^1,2, and Rahel Heule^1,2,3
1Department of High-Field Magnetic Resonance, Max Planck Institute for Biological Cybernetics, Tübingen, Germany, ²Department of Biomedical Magnetic Resonance, University of Tübingen, Tübingen, Germany, ³Center for MR Research, University Children's Hospital, Zurich, Switzerland

Keywords: Machine Learning/Artificial Intelligence, Machine Learning/Artificial Intelligence, Model-Based Learning

The bSSFP sequence is intrinsically sensitive to T1 and T2, motion robust, and allows highly efficient data acquisition. Slow convergence in qMRI parameter fitting can potentially be mitigated by machine learning, which benefits greatly from the availability of accurate ground truth data. This work presents an unsupervised model-based NN that incorporates the analytical bSSFP signal equation into the training loop, thus avoiding the need for ground truth relaxometry measurements and enabling instantaneous multi-parametric submillimeter whole-brain mapping of T1 and T2. NN performance was compared to MIRACLE quantitatively for in silico noise corrupted data and qualitatively for in vivo data.

View Presentation Video

Jeff Snyder¹, Alan H. Wilman¹, and Rob Stobbe^1
1Biomedical Engineering, University of Alberta, Edmonton, AB, Canada

Keywords: Pulse Sequence Design, Brain

An efficient k-space sampling strategy using 3D Yarnball was used in long echo train FSE (TurboYarn).  As TurboYarn samples many lines worth of k-space in each readout, fully sampled isotropic 1 mm brain scans were acquired in 40 s.  TurboYarn was compared to a time-equivalent, 4x under-sampled 3D FSE sequence in brain of three healthy subjects.  Resulting images showed excellent GM/WM contrast for TurboYarn, with CNRs higher than 3D FSE.  As each readout samples k-space centre, further work will investigate TurboYarn motion robustness, with additional SNR improvements via trajectory and flip angle train optimizations.

View Presentation Video

Xiangchuang Kong¹, Jun Ma², Erin Westerhold¹, Eric H. Middlebrooks¹, Shengzhen Tao¹, Chen Lin¹, and Xiangzhi Zhou^1
1Department of Radiology, Mayo Clinic, Jacksonville, FL, United States, ²Siemens Medical Solutions USA, Inc., Jacksonville, FL, United States

Keywords: Blood vessels, High-Field MRI, DANTE, SPACE, 7T

MR vessel-wall-imaging (VWI) is used clinically for noninvasive characterization of vessel wall pathology. The DANTE preparation module has been combined with conventional 3D-T1W-SPACE to achieve better slow-flow-suppression. 7T could further improve VWI by its increased SNR and spatial-resolution. Here we sought to optimize DANTE preparation module for T1W-SPACE at 7T. Simulations were done to determine the optimal parameters for the DANTE module. Human scans showed that the optimized T1W- DANTE-SPACE at 7T efficiently suppressed slow-blood-flow in both intracranial arteries and veins, and in turn improved wall contrast compared with conventional T1W-SPACE.

View Presentation Video

Naoyuki Takei¹, Baolian Yang², Brian Burns³, Yoichiro Ikushima¹, R Marc Lebel⁴, Vince Magnotta⁵, Fumiyasu Tsushima⁶, Shingo Kakeda⁶, Atsushi Nozaki¹, and Tetsuya Wakayama^1
1GE Healthcare, Tokyo, Japan, ²GE Healthcare, Waukesha, WI, United States, ³GE Healthcare, Menlo Park, CA, United States, ⁴GE Healthcare, Calagary, AB, Canada, ⁵University of Iowa, Iowa City, IA, United States, ⁶Hirosaki University Graduate School of Medicine, Aomori, Japan

Keywords: Blood vessels, Machine Learning/Artificial Intelligence, 7T MRI

Ultra-high-field MRA shows the promise to improve visualize the microvasculature and become an important investigation tool for research. However, increasing the consistency in image quality and reliability of small vessel visualization is necessary to demonstrate useful MR applications in clinical practice. We developed a deep learning-based reconstruction method to provide reduced noise and enhanced spatial resolution. The technique was evaluated by applying it to 3DTOF MRA data at 3T and 7T, demonstrating improved small vessel depiction.

View Presentation Video

Nicholas Farley¹, Brian Bozymski¹, Ulrike Dydak¹, and Uzay Emir^1
1Radiological Health Science, Purdue University, West Lafayette, IN, United States

Keywords: Gray Matter, Brain

To develop a reasonably fast P-31 3D MRSI sequence utilizing Rosette-Petal trajectories in K-space and compressed sensing reconstruction to acquire 1 mL voxel resolution maps in under 10 minutes with a magnetic field strength of only 3 Tesla

View Presentation Video

Nguyen Phuoc Huynh¹ and Gopikrishna Deshpande^1
1Electrical and Computer Engineering, Auburn University, Auburn, AL, United States

Keywords: Data Analysis, High-Field MRI, Effective connectivity

High-resolution fMRI presents opportunities in the study of fine-scale functional architecture, such as directional interactions between cortical layers. The Default Mode Network (DMN) is a major network in the brain associated with many social behaviors, however, little is known about the feed-forward and feedback information flow between different layers in regions of the DMN. We used high-resolution fMRI combined with effective connectivity to investigate the layer-to-layer interactions of regions in the DMN including medial prefrontal cortex, posterior cingulate cortex, and inferior parietal cortex. The results suggest that the middle layers of PCC are dominant in information flow within the DMN.

Xin Li¹, Yue Qin¹, Yanqiang Qiao¹, Yifan Qian¹, Xiaoshi Li¹, and Lei Wang^1
1Xi'an Daxing Hospital, Xi'an, China

Keywords: Stroke, fMRI (resting state)

The primary aim of the research was to compare the impact of subacute ischemic stroke on brain activity using resting state functional magnetic resonance imaging. In this study, we applied ReHo and FC to investigate the characteristics of brain functional activity in subacute ischemic stroke patients. Our results showed that the subacute ischemic stroke patients showed increased ReHo values in the left postcentral gyrus and significantly decreased FC between the Cerebellum-SUIT_7 and the right caudate, which is beneficial to understand the neurophysiological mechanisms of subacute ischemic stroke.

View Presentation Video

James Moggridge^1,2, Sjoerd Vos^1,3,4, Ferran Prados³, Matthew Grech-Sollars^1,3, Daniel Matthews⁵, Peter Schmidt⁵, Daniel C Alexander³, John S Duncan⁶, Frederik Barkhof^1,3,7, Tarek A Yousry^1,2, and John S Thornton^1,2
1Lysholm Department of Neuroradiology, National Hospital for Neurology and Neurosurgery, University College London Hospitals NHS Foundation Trust, London, United Kingdom, ²Department of Brain Repair and Rehabilitation, UCL Institute of Neurology, University College London, London, United Kingdom, ³Centre for Medical Image Computing, Department of Computer Science, University College London, London, United Kingdom, ⁴Centre for Microscopy, Characterisation, and Analysis, University of Western Australia, Perth, Australia, ⁵Advanced Research Computing Centre, University College London, London, United Kingdom, ⁶Department of Clinical and Experimental Epilepsy, UCL QS Institute of Neurology, London, United Kingdom, ⁷Department of Radiology and Nuclear Medicine, Amsterdam UMC, Amsterdam, Netherlands

Keywords: Data Analysis, Software Tools, Containerisation

The challenges encountered in development and clinical deployment of quantitative neuroradiological software techniques for in-house use can be overcome to some degree, through the application of a multi-disciplinary approach, and the use of containerisation. However, full life-cycle support including post-market surveillance presents a more significant obstacle.  

View Presentation Video

Xiaoping Wu¹, Kamil Ugurbil¹, Gregory J. Metzger¹, and Xiufeng Li^1
1University of Minnesota, Minneapolis, MN, United States

Keywords: Data Processing, Brain, ASL, cerebral perfusion or blood flow

To ensure satisfactory perfusion signal-to-noise ratio (SNR) for ASL imaging, intravascular perfusion signals are typically not suppressed (e.g., high-resolution multi-delay SMS/MB-EPI PCASL imaging for Human Connectome Projects) but can induce significant bias for CBF quantification if not addressed sufficiently. Recently proposed image denoising methods may be able to improve ASL imaging quality and SNR, having the potential to making it practical to perform ASL imaging with intravascular signal suppression. Our preliminary study results suggested that with image denoising, high-resolution multi-delay SMS/MB-EPI PCASL imaging with intravascular suppression could be achieved within 7 minutes.

View Presentation Video

Lauren F O'Donnell¹, Gonzalo Rodriguez¹, Gregory Lemberskiy¹, Zidan Yu^1,2,3, Olga Dergachyova¹, Martijn A Cloos^4,5, and Guillaume Madelin^1
1Department of Radiology, Center for Biomedical Imaging, New York University Grossman School of Medicine, New York, NY, United States, ²Department of Medicine, John A. Burns School of Medicine, University of Hawaii, Honolulu, HI, United States, ³Vilcek Institute for Graduate Biomedical Sciences, NYU Langone Health, New York, NY, United States, ⁴Centre for Advanced Imaging, The University of Queensland, Brisbane, QLD, Australia, ⁵ARC Training Centre for Innovation in Biomedical Imaging Technology, The University of Queensland, Brisbane, QLD, Australia

Keywords: MR Fingerprinting/Synthetic MR, Brain, Sodium, x - nuclei, relaxometry, 7 Tesla

 

We present a new MRF technique for simultaneous mapping of sodium T₁, biexponential T₂*, ion density, B⁺₁ shift and B₀ in the brain at 7T. This is accomplished using a 23 pulse MRF train combined with a UTE FLORET trajectory readout. Our technique is demonstrated in a 3 compartment phantom and in vivo across four healthy volunteers. Matching to a dictionary with over 800,000 signals produced parameter maps with good agreement to literature values. Flow and partial volume effects may interfere with matching T*_2,short in CSF.

View Presentation Video

Jayse Merle Weaver^1,2, Steven Kecskemeti², Jose M Guerrero-Gonzalez^1,2, Andrew L Alexander^1,2,3, and Douglas C Dean III^1,2,4
1Medical Physics, University of Wisconsin-Madison, Madison, WI, United States, ²Waisman Center, University of Wisconsin-Madison, Madison, WI, United States, ³Psychiatry, University of Wisconsin-Madison, Madison, WI, United States, ⁴Pediatrics, University of Wisconsin-Madison, Madison, WI, United States

Keywords: Quantitative Imaging, Brain, Myelin

A T1-only method to extract a quantitative myelin water fraction, utilizing a 3D radial k-space MPnRAGE sequence that reconstructs hundreds of different inversion time (TI) images in the same scan used for high-resolution T1-weighted morphometry was developed. To address high sensitivities to additive image noise of this complex model, a multi-pass fitting algorithm was prototyped that provides educated initial guesses and constraints on the underlying parameters.

Haoran Li¹, Cheng Li¹, Weijian Huang¹, Yeqi Wang^1,2, Yu Zhang¹, Xue Liu¹, Hairong Zheng¹, and Shanshan Wang^1,3,4
1Paul C. Lauterbur Research Center for Biomedical Imaging, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China, ²School of Computer and Artificial Intelligence, Zhengzhou University, Zhengzhou, China, ³Peng Cheng Laboratory, Shenzhen, China, ⁴Guangdong Provincial Key Laboratory of Artificial Intelligence in Medical Image Analysis and Application, Guangdong Provincial People’s Hospital, Guangdong Academy of Medical Sciences, Guangzhou, China

Keywords: Segmentation, Brain

Automated brain tumor segmentation with multi-modal magnetic resonance imaging (MRI) data is crucial for brain cancer diagnosis. Nevertheless, in clinical applications, it is difficult to guarantee that complete multi-modal MRI data are available due to different imaging protocols and inevitable data corruption. A large test time performance drop could happen. Here, we design a modality-adaptive network learning method to extract common representations from different modalities and make our trained model applicable to different data-missing scenarios. Experiments on an open-source dataset demonstrate that our method can reduce the dependence of deep learning-based segmentation methods on the integrity of input data.

View Presentation Video

Benjamin Lambert^1,2, Florence Forbes³, Senan Doyle², Alan Tucholka², and Michel Dojat^1
1Univ. Grenoble Alpes, Inserm, U1216, Grenoble Institut Neurosciences, GIN, Grenoble, France, ²Pixyl, Research and Development Laboratory, Grenoble, France, ³Univ. Grenoble Alpes, Inria, CNRS, Grenoble INP, LJK, Grenoble, France

Keywords: Machine Learning/Artificial Intelligence, Artifacts

The performance of Deep Learning (DL) models may drastically drop in the presence of characteristics in test images not present in the training set. Then, the automatic detection of these Out-Of-Distribution (OOD) inputs is important to deploy these methods especially for clinical applications. We address this issue in the context of DL-based subcortical structures segmentation on T1w brain MRI. We compare two OOD detection frameworks equipping DL segmentation models, Maximum Softmax Probability and Deterministic Uncertainty Method, and demonstrate the superiority of the latter which allows a robust and versatile identification of artifacts in images. 

View Presentation Video

Sarah Reeve^1,2, Alessandro Guida^2,3, Chris Bowen^1,2,3, James Rioux^1,2,3, David Volders^3,4, Jens Heidenreich^3,4, and Steven Beyea^1,3,5,6
1Physics and Atmospheric Science, Dalhousie University, Halifax, NS, Canada, ²Biomedical Translational Imaging Centre, QEII Health Sciences Centre, Halifax, NS, Canada, ³Diagnostic Radiology, Dalhousie University, Halifax, NS, Canada, ⁴Diagnostic Imaging, QEII Health Sciences Centre, Halifax, NS, Canada, ⁵Biomedical Translational Imaging Centre, IWK Health Centre, Halifax, NS, Canada, ⁶School of Biomedical Engineering, Dalhousie University, Halifax, NS, Canada

Keywords: Machine Learning/Artificial Intelligence, Head & Neck/ENT, Super Resolution

Low resolution (LR) balanced steady-state free precession (bSSFP) acquisitions confer decreased acquisition times, reduced patient motion and heating, and increased artifact tolerance due to a decrease in TR. The application of a pre-trained super resolution network to LR bSSFP images of the temporal bone allows for these advantages to be realized, without significantly degrading image quality. In the absence of a matched high resolution image, quality was judged by two radiologists. However, radiologist’s ratings were not in agreement, highlighting the fact that there is no single definition of task-specific image quality, which must be considered when super resolution is performed.