Yun Jiang¹, Ruogu Matthew Zhu², Bo Zhao^3,4, and Nicole Seiberlich^1
1Department of Radiology, University of Michigan, Ann Arbor, MI, United States, ²Department of Electrical and Computer Engineering, University of Michigan, Ann Arbor, MI, United States, ³Department of Biomedical Engineering, University of Texas at Austin, Austin, TX, United States, ⁴Oden Institute for Computational Engineering and Sciences, University of Texas at Austin, Austin, TX, United States

The purpose of this work is to assess the feasibility of performing MR Fingperprinting with 3D pseudorandom Cartesian sampling for an isotropic spatial resolution of 1mm3 at 0.55T in the brain. 

Shu-Fu Shih^1,2, Sophia X. Cui³, Xiaodong Zhong³, Bilal Tasdelen⁴, Ecrin Yagiz⁴, Krishna S. Nayak⁴, and Holden H. Wu^1,2
1Department of Radiological Sciences, University of California Los Angeles, Los Angeles, CA, United States, ²Department of Bioengineering, University of California Los Angeles, Los Angeles, CA, United States, ³Siemens Medical Solutions USA, Inc., Los Angeles, CA, United States, ⁴Ming Hsieh Department of Electrical and Computer Engineering, University of Southern California, Los Angeles, CA, United States

MRI proton-density fat fraction (PDFF) is used for non-invasive diagnosis of fatty liver disease, and has been validated at 1.5T and 3T field strengths. There is renewed interest in lower field strengths, such as 0.55T, because of potential advantages such as lower hardware/siting costs and a larger bore diameter. This study investigated PDFF quantification at 0.55T and validated PDFF quantification at 0.55T using Cartesian and Radial Dixon techniques in a reference phantom, and demonstrated the feasibility of in vivo free-breathing radial liver PDFF quantification in healthy subjects at 0.55T.

Gregory Lemberskiy^1,2, Vasiliki Mallikourti³, Dmitry S Novikov¹, and Lionel M Broche^3
1Radiology, New York University Grossman School of Medicine, New York, NY, United States, ²Microstructure Imaging INC, New York, NY, United States, ³Aberdeen Biomedical Imaging Centre, University of Aberdeen, Aberdeen, United Kingdom

We showcase high SNR T1 maps over a range of low field strengths on a stroke patient imaged using a field-cycling imaging scanner (FCI), an MR system capable of varying B0 over 2 orders of magnitude. Images from this scanner are denoised using Random Matrix Theory (RMT), leveraging the redundancy over multiple inversion times and field strengths to increase the baseline SNR by 3-fold. Following RMT denoising, we observe the T1 dispersion effect indicating a deviation of R1 frequency dependence from the BPP theory.

Krishna S. Nayak^1,2, Bilal Tasdelen¹, Ecrin Yagiz¹, and Sophia X. Cui^3
1Ming Hsieh Department of Electrical and Computer Engineering, University of Southern California, Los Angeles, CA, United States, ²Department of Biomedical Engineering, University of Southern California, Los Angeles, CA, United States, ³Siemens Medical Solutions USA, Inc., Los Angeles, CA, United States

We demonstrate single-breath-hold volumetric water/fat imaging of the abdomen using spiral gradient echo imaging at 0.55T. This provides 4x finer spatial resolution with adequate SNR and spatial resolution to measure the visceral adipose tissue compartment.

Anna Lavrova¹, Kathleen Ropella-Panagis¹, Nancy Dudek¹, Joel Morehouse¹, Ryo Kurokawa¹, Mariko Kurokawa¹, Pedro Itriago-Leon², 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

The purpose of this study was to determine which sequences can be immediately deployed for brain and spine imaging on a low-field 0.55T MRI system, with the goal of generating optimized protocols which can be used at other sites for routine clinical neuroimaging. The image quality of 174 brain and spine images was rated by two neuroradiologists. Almost all brain (T1w, T2w, FLAIR, DWI, MRA) and spine (T1w, T2w, DIXON) sequences received image quality scores corresponding to acceptable diagnostic image quality. These sequences form numerous neurological MRI protocols, which based on these results could be collected at 0.55T. 

Nathan S Artz¹, Maura S Sien¹, Amie S Robinson¹, Houchun H Hu², Rafael O'Halloran², Megan Poorman², and Sherwin S Chan^1
1Radiology, Children's Mercy Kansas City, Kansas City, MO, United States, ²Hyperfine, Inc., Guilford, CT, United States

Low-field, portable MRI offers great potential to improve diagnostic ability in the neonatal intensive care unit. Brain relaxometry will be critical to tune scan parameters at a low magnetic field for infants, whose brain tissue composition is very different than adults. In this study, ten infants were scanned at bedside on a 64mT system. Whole brain T1 and T2 mapping was performed for five infants each. Regions of interest were drawn in the white matter and thalamus for all subjects. Mean T1 and T2 values were 702ms and 294ms for white matter and 364ms and 139ms for the thalamus.

Joshua Harper¹, Venkateswararao Cherukuri², Tom O'Reilly³, Mingzhao Yu², Edith Mbabazi-Kabachelor⁴, Roland Mulando⁴, Kevin N. Sheth⁵, Andrew G. Webb³, Benjami C. Warf⁶, Abhaya V. Kulkarni⁷, Vishal Monga², and Steven J. Schiff^2
1Engineering Science, Penn State University, State College, PA, United States, ²Penn State University, University Park, PA, United States, ³Leiden University Medical Center, Leiden, Netherlands, ⁴The CURE Children's Hospital of Uganda, Mbale, Uganda, ⁵Yale, New Haven, CT, United States, ⁶Boston Children's Hospital, Harvard Medical School, Boston, MA, United States, ⁷University of Toronto, Toronto, ON, Canada

Brain images of a quality lower than is conventionally acceptable may still be useful for planning hydrocephalus treatment. Low-field MRI is a technology of growing global interest which has the capability of producing images of sufficient quality for treatment planning and is affordable enough to disseminate to rural regions of the world. Though deep learning enhancement of low-quality images does improve CNR and apparent quality, spatial errors of brain and CSF after enhanced reconstruction add significant risk to treatment management and should be avoided.

Francesco Padormo^1,2, Paul Cawley¹, Louise Dillon¹, Emer Hughes¹, Jennifer Almalbis¹, Joanna Robinson¹, Alessandra Maggioni¹, Daniel Cromb¹, Anthony Price¹, UNITY Consortium³, Lori Arlinghaus⁴, Houchun Harry Hu⁴, Steve Williams³, Shaihan Malik¹, Serena Counsell¹, Mary Rutherford¹, Tomoki Arichi¹, A. David Edwards¹, and Joseph V. Hajnal^1
1Centre for the Developing Brain, King's College London, London, United Kingdom, ²Medical Physics, Guy's & St Thomas' NHS Foundation Trust, London, United Kingdom, ³Centre for Neuroimaging Sciences, King's College London, London, United Kingdom, ⁴Hyperfine Inc., Guildford, CT, United States

A promising application of ultra-low field portable MRI is neonatal imaging. In this work we show that neonatal brain T1 mapping is viable at 64mT utilising the Hyperfine Swoop portable MRI system.

Kalina V Jordanova¹, Michele N Martin¹, Karl F Stupic¹, Samantha By², Megan E Poorman², Rafael O'Halloran², and Kathryn E Keenan^1
1NIST: National Institute of Standards and Technology, Boulder, CO, United States, ²Hyperfine, Inc., Guilford, CT, United States

Increased use of low-field MRI and quantitative MRI suggests a need for accurate quantitative measurements for fields <0.55T. We measure T₁ and T₂ for normal brain tissues at 0.064T using a Hyperfine scanner. T₁ measurements gave a mean of 266±8.4ms for white matter, 358±6.6ms for grey matter; T₂ measurements gave a mean of 95.8±2.4ms for combined white matter and grey matter, 794±357ms for CSF. Validation samples are measured on the Hyperfine system and compared to measurements taken on an electromagnet NMR at 0.064T.

Sharada Balaji¹, Adam Dvorak¹, Irene M. Vavasour^2,3, Megan E. Poorman⁴, Hanwen Liu¹, Emil Ljungberg^5,6, Steve Williams⁶, Sean Deoni⁷, Cornelia Laule^1,2,3,8, David K.B. Li², G.R. Wayne Moore^3,8,9, Alex MacKay^1,2, and Shannon H. Kolind^1,2,3,9
1Physics and Astronomy, University of British Columbia, Vancouver, BC, Canada, ²Radiology, University of British Columbia, Vancouver, BC, Canada, ³International Collaboration on Repair Discoveries, Vancouver, BC, Canada, ⁴Hyperfine, Inc., Guilford, CT, United States, ⁵Medical Radiation Physics, Lund University, Lund, Sweden, ⁶Neuroimaging, King's College London, London, United Kingdom, ⁷MNCH D&T, Bill and Melinda Gates Foundation, Seattle, WA, United States, ⁸Pathology and Laboratory Medicine, University of British Columbia, Vancouver, BC, Canada, ⁹Medicine, University of British Columbia, Vancouver, BC, Canada

A single hemisphere of postmortem brain tissue from a subject with multiple sclerosis (MS) was scanned at high (Philips Elition 3T) and low (Hyperfine Swoop 64mT) field strength to determine whether lesions could be detected using a portable low field MRI system. T₂-weighted scans were acquired at both field strengths with matching resolutions to assess the impact of low field. Comparisons with a high resolution 3T scan showed that of 17  visible lesions, 11 were seen on the lower resolution 3T and 10 were seen on the 64mT scan, demonstrating the feasibility of lesion detection at low field.   

David Schote¹, Lukas Winter¹, Christoph Kolbitsch¹, Felix Zimmermann¹, Thomas O'Reilly², Andrew Webb², Frank Seifert¹, and Andreas Kofler^1
1Physikalisch-Technische Bundesastalt (PTB), Braunschweig and Berlin, Germany, ²C.J. Gorter Center for High Field MRI, Department of Radiology, Leiden University Medical Center, Leiden, Netherlands

A physics-informed deep neural network using a U-Net in combination with a multifrequency interpolation is presented to correct for B₀-field image distortions in low-field MRI. Training data is based on 1.5T and 3.0T knee images and realistic measurement-based assumption of SNR and B₀-field inhomogeneities from previously constructed Halbach magnets. Significant (p<0.05) improvements are demonstrated applying the suggested methodology to uncorrected images. 

M. Ethan MacDonald^1,2, Eremiahs Fikre^1,2, Fernando Vega^1,2, and AbdolJalil Addeh^1,2
1Biomedical Engineering, Electrical & Software Engineering, Radiology, University of Calgary, Calgary, AB, Canada, ²Hotchkiss Brain Institute, Foothills Medical Centre, Calgary, AB, Canada

In this work, the denoising autoencoder is applied to simulated low field, low resolution, low signal-to-noise images and used to recover the high field, high resolution, high signal-to-noise paired image.  Different types of noise, gaussian and chi-squared, is added in simulation.  We found that the denoising autoencoder worked slightly better for normally disturbed noise, but not in all cases.  We found a linear trend between the model performance with RMSE and the standard deviation of the added noise.  This work demonstrates the use of simple and robust denoising autoencoder to improve low field MRI. 

David E J Waddington¹, Efrat Shimron², Nicholas Hindley^1,3, Neha Koonjoo³, and Matthew S Rosen^3,4,5
1ACRF Image X Institute, Faculty of Medicine and Health, The University of Sydney, Sydney, Australia, ²2Department of Electrical Engineering and Computer Sciences, UC Berkeley, Berkeley, CA, United States, ³A. A. Martinos Center for Biomedical Imaging, Charlestown, MA, United States, ⁴Department of Physics, Harvard University, Cambridge, MA, United States, ⁵Harvard Medical School, Boston, MA, United States

Portable MRI scanners that operate at very low magnetic fields are increasingly being deployed in clinical settings. However, the intrinsic low signal-to-noise (SNR) ratio of these low-field MRI scanners often necessitates many signal averages, and therefore excessively long acquisition times. Here we propose to improve SNR through optimized k-space undersampling and Compressed Sensing reconstruction. We demonstrate this approach for 6.5 mT ultra-low-field MRI using: (1) retrospective-subsampling experiments with 2x to 4x acceleration; (2) prospectively-subsampled data acquired from a human brain phantom with a 6.5mT MRI. The results exhibit a higher SNR than the traditional averaging method, without increasing scan time.

Fernando Galve^1,2, José Miguel Algarín^1,2, Teresa Guallart³, Rubén Pellicer⁴, Eduardo Pallás^1,2, José Manuel González³, Yolanda Vives³, Rubén Bosch³, Guillermo López^1,2, Juan Pablo Rigla³, José Borreguero³, Alfonso Ríos³, José María Benlloch^1,2, and Joseba Alonso^1,2
1MRILab, Institute for Molecular Imaging and Instrumentation (i3M), Spanish National Research Council (CSIC), Valencia, Spain, ²MRILab, Institute for Molecular Imaging and Instrumentation (i3M), Universitat Politècnica de València (UPV), Valencia, Spain, ³Tesoro Imaging SL, Valencia, Spain, ⁴Physio MRI SL, Valencia, Spain

Here we present accelerated images by PhasE-Constrained Over Sampled (PECOS) MRI in our home-made Halbach MRI scanner (70 mT). We show projection images of a phantom obtained in our “PhysioMRI” scanner (Fig. 1(a)), a home-made scanner designed for musculoskeletal imaging at low field strengths. We use cartesian turbo spin echo (TSE) sequence to sample the k-space. Image reconstructions are performed by phase conjugate reconstruction and by PECOS.

Sara Neves Silva¹, Irina Grigorescu², Alena Uus², Johannes Steinweg², Maria Deprez², Jo Hajnal², Kuberan Pushparajah², Jana Hutter², and Enrico De Vita^2
1Biomedical Engineering and Imaging Sciences, King's College London, London, United Kingdom, ²King's College London, London, United Kingdom

Fetal MRI and MRS are often compromised due to unpredictable fetal motion and commonly require multiple repetitions for diagnostic studies. To overcome these limitations, we are working towards developing a deep-learning based automatic MRI fetal motion tracking method. Our method uses, as input, rapid dynamic multi-echo 2D-EPI acquisitions and is based on a 3D U-Net for brain localisation and translational motion parameters estimation for brain tracking. The results show rapid low-resolution acquisitions contain sufficient information to allow automated fetal brain localisation. Our technique can be used to assess fetal movements and to build navigation systems for fetal MRI/MRS.