View Presentation Video

Lucie Chalet^1,2, Timothé Boutelier², Thomas Christen³, Justine Debatisse⁴, Oceane Wateau⁵, Nicolas Costes⁶, Ines Merida⁶, Sophie Lancelot^6,7, Christelle Leon¹, Norbert Nighoghossian^1,8, Yves Berthezene^9,10, Omer Faruk Eker^9,10, Tae-Hee Cho^1,8, Emmanuelle Canet-Soulas¹, and Laura Mechtouff^1,8
1Univ Lyon, CarMeN Laboratory, INSERM, INRA, INSA Lyon, Universite Claude Bernard Lyon 1, Lyon, France, ²Olea Medical, La Ciotat, France, ³Univ. Grenoble Alpes, Inserm, U1216, Grenoble Institut Neurosciences, Grenoble, France, ⁴Institut des Sciences Cognitives Marc Jeannerod (ISCMJ), Unite Mixte de Recherche 5229 du Centre National de la Recherche Scientifique (CNRS), Bron, France, ⁵Cynbiose SAS, Marcy-L'Etoile, France, ⁶CERMEP - Imagerie du Vivant, Lyon, France, ⁷Hospices Civils of Lyon, Lyon, France, ⁸Stroke department, Hospices Civils of Lyon, Lyon, France, ⁹CREATIS, CNRS UMR-5220, INSERM U1206, Universite Lyon 1, INSA Lyon, Villeurbanne, France, ¹⁰Neuroradiology department, Hospices Civils of Lyon, Lyon, France

Keywords: Data Processing, Oxygenation, Oxygen Metabolism

Latest challenges in acute ischemic stroke management involve tissue-based approaches to define a tailored ischemic penumbra to each patients. Oxygen metabolism imaging offers new perspectives to tackle these challenges. A fully automated pipeline was developed providing oxygen metabolism parameters from MRI acquisitions including advanced processing methods such as motion-correction, registration and bayesian-based deconvolution. The method was tested on an ischemia-reperfusion non-human primate model with simultaneous PET-MRI acquisitions providing gold-standard comparison of oxygen parameters. Results show strong similarities in ratio of penumbral tissues detected in the ischemic hemisphere between the two modalities and opens new perspectives of improvement in parameters definition.

View Presentation Video

Andrew N. Van^1,2, M. Dylan Tisdall³, Timothy O. Laumann⁴, David F. Montez^2,4, Damien A. Fair^5,6,7, and Nico U.F. Dosenbach^1,2,8,9
1Department of Biomedical Engineering, Washington University in St. Louis, St. Louis, MO, United States, ²Department of Neurology, Washington University School of Medicine, St. Louis, MO, United States, ³Department of Radiology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA, United States, ⁴Department of Psychiatry, Washington University School of Medicine, St. Louis, MO, United States, ⁵Institute of Child Development, University of Minnesota Medical School, Minneapolis, MN, United States, ⁶Department of Pediatrics, University of Minnesota Medical School, Minneapolis, MN, United States, ⁷Masonic Institute for the Developing Brain, University of Minnesota Medical School, Minneapolis, MN, United States, ⁸Department of Radiology, Washington University School of Medicine, St. Louis, MO, United States, ⁹Department of Pediatrics, Washington University School of Medicine, St. Louis, MO, United States

Keywords: Data Processing, fMRI

Echo Planar Imaging (EPI) suffers from off-resonance image distortion due to B0 inhomogeneities. These distortions are typically corrected by separate field map acquisitions, but only allow for sparse temporal sampling of B0 inhomogeneity. This approach prevents distortion correction if changes in the B0 field occur over time due to motion. Here, we present a method of correcting off-resonance distortion by computing a time series of field maps using phase information from multi-echo (ME) EPI data. We show that our method produces corrections comparable to the current standard of distortion correction, in addition to enabling frame-by-frame correction in functional MRI data.

Tobias C Wood¹ and Emil Ljungberg^1,2
1King's College London, London, United Kingdom, ²Lund University, Lund, Sweden

Keywords: Image Reconstruction, Image Reconstruction

We show that multiple clinical contrasts can be obtained from silent Zero Echo Time scans by using preparation pulses and a physics-informed subspace "T1-shuffling" reconstruction approach. This models and removes the effect of T1-recovery during the readout segment, which would otherwise wash out the desired contrast.

View Presentation Video

Eric A. Borisch¹, Adam T Froemming¹, Roger C. Grimm¹, Daniel Reiter², Phillip J. Rossman¹, and Stephen J. Riederer^1
1Radiology, Mayo Clinic, Rochester, MN, United States, ²St. Olaf College, Northfield, MN, United States

Keywords: Motion Correction, Motion Correction

The addition of k_y≈0 navigator views to detect anterior-posterior prostate (peristalsis-induced) motion to the end of each echo train for a 2D T2-weighted acquisition stack with slice-profile recovery reconstruction is investigated. ROVir virtual coils are used to focus the spatial sensitivity in the phase direction of the navigator response on the small, central prostate region. Mutual information is used to register these navigators across the acquisition, with the resulting movement time course used to correct the acquired data. Results both in motion controlled phantom and human studies are discussed.

View Presentation Video

James L. Kent¹, Ladislav Valkovič^2,3, Iulius Dragonu⁴, Mark Chiew^1,5,6, and Aaron T. Hess^1
1Wellcome Centre for Integrative Neuroimaging, FMRIB, Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom, ²Oxford Centre for Clinical Magnetic Resonance Research, John Radcliffe Hospital, University of Oxford, Oxford, United Kingdom, ³Department of Imaging Methods, Institute of Measurement Science, Slovak Academy of Sciences, Bratislava, Slovakia, ⁴Research & Collaborations GB&I, Siemens Healthcare Ltd, Camberley, United Kingdom, ⁵Department of Medical Biophysics, University of Toronto, Toronto, ON, Canada, ⁶Physical Sciences, Sunnybrook Research Institute, Toronto, ON, Canada

Keywords: Motion Correction, Motion Correction

RF sensor methods for motion correction require no (or minimal) additional hardware, are sequence independent and have high temporal resolution. Several RF sensor-based methods for rigid-body head motion detection have been demonstrated but which method offers the most sensitivity to motion is yet unknown. We aim to compare the sensitivity of PT and pTxS methods by simultaneously measuring these signals during continuous motion, training a linear model from EPI registered images and analysing the ability to predict rigid head positions. Currently, we see little difference in their ability to predict rigid head motion but further investigation is needed.

View Presentation Video

Helge J. Zöllner^1,2, Kelley M. Swanberg³, John LaMaster⁴, Antonia Kaiser⁵, Jamie Near⁶, Candace Fleischer^7,8, Brian J. Soher⁹, William T. Clarke¹⁰, and Georg Oeltzschner^1,2
1The Russell H. Morgan Department of Radiology and Radiological Science, The Johns Hopkins University School of Medicine, Baltimore, MD, United States, ²Kennedy Krieger Institute, F. M. Kirby Research Center for Functional Brain Imaging, Baltimore, MD, United States, ³Department of Biomedical Engineering, Columbia University Fu Foundation School of Engineering and Applied Science, New York, NY, United States, ⁴Faculty of Computer Science, Technische Universität München, München, Germany, ⁵Department of Radiology and Nuclear Medicine, Amsterdam Neuroscience, Amsterdam UMC, University of Amsterdam, Amsterdam, Netherlands, ⁶Physical Sciences Platform, Sunnybrook Research Institute, Toronto, ON, Canada, ⁷Department of Radiology and Imaging Sciences, Emory University School of Medicine, Atlanta, GA, United States, ⁸Department of Biomedical Engineering, Georgia Institute of Technology and Emory University, Atlanta, GA, United States, ⁹Center for Advanced MR Development, Department of Radiology, Duke University Medical Center, Durham, NC, United States, ¹⁰Wellcome Centre for Integrative Neuroimaging, FMRIB, Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, United Kingdom

Keywords: Data Analysis, Spectroscopy, open source

We propose a standardized format for storing basis sets and linear-combination modeling results. Conventions, guidelines, and conversion tools are currently developed in collaboration with the MRS community. This standardization effort will further improve data sharing and reproducibility, allowing for easier integration of MRS into larger neuroimaging frameworks.

View Presentation Video

Agnieszka Sierhej^1,2, Matt G Hall^1,2, Nadia AS Smith², and Chris A Clark^1
1University College London, London, United Kingdom, ²National Physical Laboratory, London, United Kingdom

Keywords: Segmentation, Brain, Uncertainty

Segmentation is crucial in analysing volumes of brain tissue compartments in studies of healthy development and pathological changes. Probabilistic segmentation is widely used, but the choice of threshold on probability maps has large effects on the resulting tissue volumes, and there is no consensus on optimal choice of such threshold. This work presents a way to minimise this effect by quantifying the uncertainty using Shannon entropy. Entropy map contains information about uncertainties in each individual voxel and this can be used to inform about uncertainty in MRI-based tissue segmentation.

Jie Huang^1
1Michigan State University, East Lansing, MI, United States

Keywords: Data Analysis, fMRI

The relationship between the activity of brain areas and that of the entire brain remains unknown. The temporal correlation of the BOLD time signal of an area with that of every point in the brain yields a full spatial map that characterizes the entire brain’s functional co-activity (FC) relative to that area’s activity. Analyzing the temporal correlation of the signal time courses of two areas and the spatial correlation of their corresponding two FC maps for all pairwise areas revealed a quantitative relationship between the activity of brain areas and that of the entire brain.

Mahshid Soleymani¹, Yunyan Zhang², and Mariana Bento^2
1Biomedical Engineering, University of Calgary, Calgary, AB, Canada, ²University of Calgary, Calgary, AB, Canada

Keywords: Machine Learning/Artificial Intelligence, Data Processing

Disease activity varies between patients with multiple sclerosis (MS), and patients who have a greater risk of developing a progressive course require more aggressive therapies earlier. However, differentiating disease severity is challenging using conventional methods as the disease often progresses silently. By taking advantage of one of the most advanced quantitative methods, convolutional neural networks, we aim to develop a new deep learning model to differentiate two common MS subtypes: relapsing-remitting course from secondary progressive phenotype. This study focuses on varying image pre-processing techniques and using different data views using conventional brain MRI.

View Presentation Video

Nimra Naeem ¹, Omair Inam ², Abdul Basit ², Hammad Omer², and Tayaba Gul^2
1Electrical and Computer Engineering, COMSATS University Islamabad, Islamabad, Pakistan, ²COMSATS University Islamabad, Islamabad, Pakistan

Keywords: Parallel Imaging, Cardiovascular

The random projection method reduces the dimensionality of the data to provide attractive computational advantages in the collection and processing of high-dimensional signals. In literature, it has been successfully applied for the pMRI method i.e., GRAPPA. This paper introduces a very sparse random projection matrix with the same statistical efficiency as dense matrix. The proposed method is implemented on FPGA (working with on-chip processor) device to speed up the GRAPPA reconstruction process. The reconstruction results for in vivo 30 channel human cardiac data set show 6x speed up with little loss in accuracy. 

View Presentation Video

Philip F. Durham¹, Humberto Monsivais¹, Gianna Nossa¹, Daniel Foti², Xiaopeng Zhou³, and Ulrike Dydak^1,4
1College of Health and Human Sciences, Purdue University, West Lafayette, IN, United States, ²Department of Psychological Sciences, Purdue University, West Lafayette, IN, United States, ³Purdue Life Sciences MRI Facility, Purdue University, West Lafayette, IN, United States, ⁴Department of Radiology and Imaging Sciences, Indiana University School of Medicine, Indianapolis, IN, United States

Keywords: Data Acquisition, Brain, Glutamate, MRS

This meta-analysis discusses the intra- and inter-subject variability of glutamate measurements in MRS, comparing widely used sequences on clinical scanners (PRESS, sLASER, MEGA-PRESS, MEGA-sLASER, HERMES). To minimize variance in glutamate measurements, the authors recommend sLASER over PRESS, non-edited over edited MRS, and OFF over DIFF spectra. 

View Presentation Video

Brian Bozymski¹, Nathan Ooms¹, and Ulrike Dydak^1,2
1School of Health Sciences, Purdue University, West Lafayette, IN, United States, ²Department of Radiology and Imaging Sciences, Indiana University School of Medicine, Indianapolis, IN, United States

Keywords: Data Acquisition, Liver, Fatty Liver Disease

Optimizing a clinically feasible ³¹P liver MRSI protocol on a 3T SIEMENS system through the lens of a diffuse fatty liver disease study.  Protocol practicality is driven primarily by scan length, patient comfort, and data quality.  Voluntary scanning subjects, ranging from healthy to overweight, are sourced from the university population.  Both a dual-tuned, 8-channel ¹H/³¹P phased array coil (wrapped around the whole torso) and an 11-cm-diameter flexible surface coil are available for use.  Techniques, testing results, and relative advantages of either are discussed in detail, as well as the implications for each choice in a clinical setting.

Antonia Susnjar¹, Antonia Susnjar², Gianna Nossa³, and Ulrike Dydak^3,4
1Biomedical Engineering, Purdue University, West Lafayette, IN, United States, ²Purdue University, West Lafayette, IN, United States, ³School of Health Sciences, Purdue University, West Lafayette, IN, United States, ⁴Department of Radiology and Imaging Sciences, Indiana School of Medicine, Indianapolis, IN, United States

Keywords: Data Acquisition, Data Acquisition

To overcome acquisition downfalls of HERMES edited magnetic resonance spectroscopy at 3T, we have conducted critical optimization for four clinically relevant brain regions in neurological disorders by finding the optimal voxel rotation that affects gradient order to enhance data quality and reproducibility.  

View Presentation Video

Xuetong Zhou^1,2, Brian A. Hargreaves^1,2,3, and Philip K. Lee^2
1Department of Bioengineering, Stanford University, Stanford, CA, United States, ²Department of Radiology, Stanford University, Stanford, CA, United States, ³Department of Electrical Engineering, Stanford University, Stanford, CA, United States

Keywords: New Signal Preparation Schemes, Data Acquisition

The performance of conventional T2 or diffusion preparation using linear phase pulses or adiabatic pulses is limited by the B₁⁺ and B₀ field inhomogeneities, particularly relevant in body imaging. In this work, we proposed a numerical optimization strategy to design RF pulse pairs for non-selective T2 or diffusion preparation that perform robustly over a wide range of B₁⁺ and B₀ variations. Phantom and in vivo experiments compared the performance of preparation using the optimized pulses to that of linear-phase and adiabatic pulses.

Chen Chu¹, Jie Meng¹, Huayong Zhang², Qianqian Feng¹, Weibo Chen³, Jian He¹, Lingyun Sun², and Zhengyang Zhou^1
1Department of Radiology, Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School, Nanjing, China, ²Department of Rheumatology, Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School, Nanjing, China, ³Philips Healthcare, Shanghai, China, Shanghai, China

Keywords: Radiomics, Diffusion/other diffusion imaging techniques

The study explored more promising radiomics extracted from apparent diffusion coefficient  maps for diagnosing Sjogren’s syndrome (SS) without head＆neck MR morphology changes. A total of 119 consecutive SS participants and 95 healthy volunteers were prospectively analyzed by 3.0 T MR including diffusion weighted imaging. Forty-five radiomic parameters were selected and twenty-two radiomic parameters showed significant difference between SS and controls, in which 11 parameters had an area under the ROC curve (AUC) greater than 0.700. The SVM classification model differentiated SS from healthy controls with an AUC of 0.932 and 0.911 in the training and testing sets, respectively.

Judith Mizrachi¹, Marco Barbieri¹, Akshay Chaudhari¹, and Garry Gold^1
1Radiology, Stanford University, Stanford, CA, United States

Keywords: Hybrid & Novel Systems Technology, Data Processing, super-resolution, high resolution, Fingerprinting

We introduce a novel approach, Super-Resolution Magnetic Resonance Fingerprinting (SR-MRF), which is a combination of an optical super-resolution microscopy method, Super-resolution Optical Fluctuation Imaging (SOFI) and MR-Fingerprinting. Our preliminary results in simulated MRF data demonstrate both 2-fold resolution improvement, as demonstrated by point spread function (PSF) analysis, and tissue-type segmentation. This technique represents an original crossover between the fields of MR-imaging and super-resolution microscopy, and lays the groundwork for other such techniques to follow.

View Presentation Video

Maurizio Bergamino¹, Elizabeth G Keeling^1,2, Sydney Y Schaefer², Anna Burke¹, George Prigatano¹, and Ashley M Stokes^1
1Barrow Neurological Institute, Phoenix, AZ, United States, ²Arizona State University, Tempe, AZ, United States

Keywords: Data Analysis, Diffusion/other diffusion imaging techniques

The objective of this study was to assess white matter microstructural differences between groups of healthy controls (HC), subjective memory complaints (SMC), and cognitive impairment (CI) through multi-shell free-water (FW) corrected DTI (FW-DTI) metrics and by the peak width of skeletonized mean diffusivity (PSMD). Additionally, brain correlations at the voxel level between a functional motor measure (two-bean transfer) and DTI metrics/PSMD were explored.

View Presentation Video

Karim Fathy¹, Felix Dumais², Samantha Côté ¹, Blaise Frederick³, and Kevin Whittingstall^4
1Departement of Biomedical Imaging and Radiation Science, Universite de Sherbrooke, Sherbrooke, QC, Canada, ²Departement of Computer Science, Universite de Sherbrooke, Sherbrooke, QC, Canada, ³Departement of Psychiatry, Harvard Medical School, Belmont, MA, United States, ⁴Departement of Radiology, Universite de Sherbrooke, Sherbrooke, QC, Canada

Keywords: Machine Learning/Artificial Intelligence, Neuro, Software Tool

Currently, there is no accurate fully automated multi-class method to segment the whole cerebral arterial tree in time-of-flight magnetic resonance angiography (TOF-MRA). We developed an artificial intelligence based software tool to identify cerebral arteries in TOF-MRAs. We trained a neural network on a TOF-MRA dataset and labeled the cerebral arterial tree using different image processing techniques. Our software tool is fast and reliable, with no human intervention, and allows for the conduction of large-scale TOF-MRA studies while being versatile in segmenting a diverse set of TOF-MRAs. 

View Presentation Video

Ying Chu¹ and Jürgen Finsterbusch^1
1Department of Systems Neuroscience, University Medical Center Hamburg-Eppendorf, Hamburg, Germany

Keywords: Motion Correction, fMRI

Motion correction is crucial for functional neuroimaging, however, not straightforward for applications in the cervical spinal cord because its shape, size, and inner structure barely vary along the cord’s axis. Previous spinal cord fMRI studies involved motion correction in this direction based on the vertebral disks visible in the fMRI acquisitions but for transverse slices, this approach suffers from a low spatial resolution (slice thickness typically 3.5mm or more). Here, a columnar navigator positioned along the spine is used to detect spine movements with a high spatial resolution that could be used for motion correction of spinal cord fMRI.

View Presentation Video

Antti Paajanen¹, Olli Nykänen¹, Ville Kolehmainen¹, and Mikko J. Nissi^1
1Department of Applied Physics, University of Eastern Finland, Kuopio, Finland

Keywords: Sparse & Low-Rank Models, Quantitative Imaging, Image reconstruction

Quantitative MRI offers unique opportunities for compressed sensing reconstructions because the signal evolution is known. Here we compare a principal component subspace reconstruction to a standard total-variation regularized compressed sensing approach with a 3-D radial variable flip angle simulation data. The simulation data allows us to measure only the effect of the chosen reconstruction. The subspace approach consistently yields similar or better image quality and does not seem to require contrast dimension regularization to achieve it.

View Presentation Video

Jeff Kershaw¹ and Takayuki Obata^1
1Applied MRI Research, QST, Chiba, Japan

Keywords: Signal Representations, Diffusion/other diffusion imaging techniques, OGSE

The target of oscillating-gradient spin-echo diffusion-weighted imaging (OGSE-DWI) is the spectral density of molecular diffusion, $$$u_2(\omega)$$$, which is predicted to obey a set of asymptotic universality relations that are linked to the global organisation of the sample. In principle, the complex microstructure of a medium can be classified by measuring the spectral density in its low- and high-frequency limits. However, the quantity most often measured with OGSE-DWI is the apparent diffusion coefficient (ADC). It was the purpose of this study to investigate how the low-frequency asymptotic behaviour of the spectral density is reflected in the ADC.

View Presentation Video

Aaron T Gudmundson^1,2, Kathleen E Hupfeld^2,3, Yulu Song^2,4, Helge J Zöllner^1,2, Christopher W. Davies- Jenkins^1,2, İpek Özdemir^1,2, Georg Oeltzschner^2,3, and Richard A E Edden^1,2
1The Russell H. Morgan Department of Radiology and Radiological Sciences, Johns Hopkins University School of Medicine, Baltimore, MD, United States, ²F.M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, MD, United States, ³The Russell H. Morgan Department of Radiology and Radiological SciencesRadiology and Radiological Sciences, Johns Hopkins University School of Medicine, Baltimore, MD, United States, ⁴The Russell H. Morgan Department of Radiology and Radiological Sciences Radiology and Radiological Sciences, Johns Hopkins University School of Medicine, Baltimore, MD, United States

Keywords: Machine Learning/Artificial Intelligence, Spectroscopy, Out-of-voxel (OOV), MRS, Artefacts, Deep Learning, Convolutional Neural Network

Out-of-voxel (OOV) artefacts, or echoes, are common in-vivo artefacts seen in MRS data. These artefacts are typically not identified until post-processing and are challenging to remove without modifying the underlying data. Here, we developed 2 Convolutional Neural Networks (CNNs) to overcome OOV artefacts at different stages. The first network (CNN1) was designed to identify OOV artefacts in single average data and offer a real-time assessment during data acquisition. The second (CNN2) predicts the OOV artefact to subtract during post-processing without impacting the metabolite data.