Gaurav Verma¹, Rebecca Emily Feldman², and Priti Balchandani^1
1Radiology, Icahn School of Medicine at Mount Sinai, New York, NY, United States, ²Medical Physics, University of British Columbia, Kelowna, BC, Canada

An echo-planar spectroscopic imaging sequence has been demonstrated incorporating semi-adiabatic band-limited refocusing (SABRE) pulses. The sequence simultaneously addresses two major obstacles to the implementation of high-resolution spectroscopic imaging at ultrahigh field: presence of B₁ field inhomogeneity and aliasing due to bandwidth limitation in the spectral-spatial readout. Both bi-polar and flyback variants of the EPSI readout have been implemented in a user-selectable manner. An accompanying Matlab-based reconstruction has been developed which performs extraction and complete processing from raw data including eddy current compensation, phase reversal of even echoes in the bi-polar readout and removal of spurious points acquired during gradient switching.

Xiaoxuan He¹, Edward J. Auerbach¹, Michael Garwood¹, Naoharu Kobyashi¹, Alireza Sadeghi‐Tarakameh¹, Yigitcan Eryaman¹, Xiaoping Wu¹, and Gregory J. Metzger^1
1Center for Magnetic Resonance Research, University of Minnesota, Minneapolis, MN, United States

A parallel transmit optimized 3D spatial-spectral pulse is developed for spectroscopic imaging for brain studies at 10.5T with reduced SAR, intrinsic water suppression and field inhomogeneity mitigation. Phantom studies are used to compare the new method with a conventional approach.

Guodong Weng¹, Sulaiman Sheriff², Claus Kiefer¹, Irena Zubak³, Andrew A Maudsley², and Johannes Slotboom^1
1Institute for Diagnostic and Interventional Neuroradiology, Support Center for Advanced Neuroimaging (SCAN), University of Bern, Bern, Switzerland, ²Department of Radiology, University of Miami School of Medicine, Miami, FL, United States, ³Inselspital Bern and University Hospital, Bern, Switzerland

To tackle the chemical shift displacement artifacts (CSDA), B1-inhomoheneity, water and fat suppression as well as  specific absorption-rate (SAR) limitation at 7T spectroscopy, an EPSI-variant sequence is developed in this study. The conventional slice-selective refocusing pulse is replaced by a spectral-selective adiabatic 2π-pulse pair. The results show a 77% reduction of CSDA, homogeneous refocusing map, water suppression factor ≥ 1000 within an acceptable SAR both in vitro and in vivo measurement. The scan time for the whole brain is within 8 minutes. This new EPSI-variant sequence shows its high application potential clinical routine.

Huawei Liu¹, Adam Autry¹, Duan Xu¹, Peder Larson¹, and Yan Li^1
1Department of Radiology and Biomedical Imaging, University of California, San Francisco, San Francisco, CA, United States

Proton MR spectroscopy has been widely used for non-invasive assessment of brain metabolites. Metabolic alterations in the motor and sensory cortex, cingulate cortex and subcortical structures have been associated with pathology. It is desirable to acquire MRSI in these regions simultaneously with short scan time and sufficient SNR. We have developed a novel technique for simultaneous multiband 2D MRSI acquisition, which combines automated atlas-based prescription and adaptive Hadamard pulses. This technique helps to reduce prescription time, improve reproducibility and increase SNR efficiency.

Masoumeh Dehghani^1,2 and Jamie Near^1,2
1McGill university, Montreal, QC, Canada, ²Centre d'Imagerie Cérébrale, Douglas Mental Health University, Montreal, QC, Canada

Previously, we demonstrated the feasibility of performing simultaneous MRS localization in two human brain regions using the dual-SPECIAL technique. Based on the SPin ECho, full Intensity Acquired Localized sequence, this approach yields simultaneous acquisition from two regions in combination with Hadamard encoding, reducing the acquisition time by half compared with serial acquisition. Here, we demonstrate that MRS acquisition using dual-Special sequence reveals significant differences in the ratio of tCho/tCr between anterior and posterior regions and higher concentration of several metabolites in regions with higher gray matter fraction.      

Adam Berrington¹, Joseph S Gillen^2,3, and Vincent Boer^4
1Sir Peter Mansfield Imaging Centre, School of Physics and Astronomy, University of Nottingham, Nottingham, United Kingdom, ²Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine, Baltimore, United Kingdom, ³F.M. Kirby Research Centre, Kennedy Krieger Institute, Baltimore, MD, United States, ⁴Danish Research Centre for Magnetic Resonance,Centre for Functional and Diagnostic Imaging and Research, Copenhagen University Hospital, Hvidovre, Denmark

Gradient modulated pulses allow for accurate slice localisation in single-voxel MRS. However, they suffer from ‘smearing’ artifacts at off-resonance leading to poor performance across the inversion band, particularly when B₁ field strength is low. An optimized GOIA-WURST pulse shape was found by simulating changes in the modulation parameters to improve the off-resonance profile at 7T for B₁ of 15 μT. When used in a semi-LASER localisation, the optimised GOIA(50,10) pulses resulted in a minimum TE of 23 ms which led to upright peak shapes for J-coupled multiplets in phantom.

Mark Mikkelsen^1,2, Steve C. N. Hui^1,2, and Richard A. E. Edden^1,2
1Russell H. Morgan Department of Radiology and Radiological Science, The 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

Offsets in B₀ field frequency have a detrimental impact on the SNR, linewidth, and editing efficiency of edited ¹H MRS data. We have previously described an interleaved water referencing (IWR) method for interspersed acquisition of the unsuppressed water signal to correct frequency drift at regular intervals. Here, we add to IWR a spectral registration (SR) approach for updating the center frequency every TR. Proof-of-concept results from phantom and in vivo experiments demonstrate that SR can be successfully combined with IWR for fully prospective correction of frequency offsets across the whole edited MRS scan.

Guodong Weng¹, Claus Kiefer¹, Irena Zubak², and Johannes Slotboom^1
1Institute for Diagnostic and Interventional Neuroradiology, Support Center for Advanced Neuroimaging (SCAN), University of Bern, Bern, Switzerland, ²Department of Neurosurgery, Inselspital Bern and University Hospital, Bern, Switzerland

To overcome the B₁⁺ inhomogeneities and SAR limitation for using MRSI-editing at UHT (≥7T), a new spectral editing scheme was developed within an EPSI sequence for whole brain MRSI. The slice-selective refocusing pulse was replaced by a chemical-selective adiabatic 2π-refocusing pulse pair with varying passbands. The results show an excellent glutamate, 2HG and GABA-editing pattern with in vitro and in vivo measurements. This study has shown the excellent performance of the proposed new spectral editing scheme, and proved that it could be an alternative method for MEGA editing.

Sofie Tapper^1,2, Muhammad G. Saleh³, Helge J. Zöllner^1,2, Steve C.N. Hui^1,2, and Richard A.E. Edden^1,2
1Russell H. Morgan Department of Radiology and Radiological Science, 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, ³Department of Diagnostic Radiology and Nuclear Medicine, University of Maryland School of Medicine, Baltimore, MD, United States

The aim of this work was to investigate an alternative MRS approach for GABA editing. By performing a range of simulations, we found a long-TE scheme that reduces the influence of co-edited macromolecular signal without lower GABA losses than long-TE MEGA-PRESS. We termed this scheme ‘Mixed flip angle’ (MFA) editing, and validated the scheme with phantom measurements and a proof-of-principle in vivo measurement. Preliminary results show that longer-TE editing of GABA has the potential to increase selectivity by allowing for longer editing pulse duration, avoiding the need for symmetrical suppression of MM signals.

Kimberly Chan¹, Andreas Hock², Richard Edden^3,4, Erin MacMillan⁵, and Anke Henning^1,6
1The University of Texas Southwestern, Dallas, TX, United States, ²MR Clinical Science, Philips Health Systems, Horgen, Switzerland, ³Russell H. Morgan Department of Radiology and Radiological Science, The Johns Hopkins University School of Medicine, Baltimore, MD, United States, ⁴. M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, MD, United States, ⁵UBC MRI Research Centre, University of British Columbia, Vancouver, BC, Canada, ⁶Max Planck Institute for Biological Cybernetics, Tübingen, Germany

Macromolecule-suppressed GABA-editing with symmetrical suppression is often preferred over conventional GABA-editing due to its greater specificity.  However, this pulse sequence is more sensitive to magnetic field instabilities than conventional GABA-editing.  This leads to macromolecule contamination in the edited GABA signal.  Here, we combine metabolite cycling with J-difference (MC-MEGA) editing to allow for prospective volume-localized frequency correction at each repetition time without the acquisition of additional water reference transients.  We show here that prospective MC-MEGA reduces B0 field instability relative to intermittent prospective frequency correction with water suppressed (WS) MEGA and reduces macromolecule contamination and subtraction artifacts.

Martyna Dziadosz¹, Maike Hoefemann¹, André Döring², Malgorzata Marjanska³, Edward Auerbach³, and Roland Kreis^1
1Departments of Radiology and Biomedical Research, University of Bern, Bern, Switzerland, ²Cardiff University Brain Research Imaging Centre (CUBRIC), School of Psychology, Cardiff University, Cardiff, United Kingdom, ³Department of Radiology, Center for Magnetic Resonance Research, University of Minnesota, Minnesota, MN, United States

The detection and quantification of Nicotinamide adenine dinucleotide (NAD⁺) was demonstrated to be challenging in ¹H Magnetic Resonance Spectroscopy, due to its low concentration and reported polarization exchange with water. Frequency-selective excitation with slice-selective refocusing was proposed to prevent saturation-transfer from water. In this study we compare three techniques to access NAD⁺ quantification – standard WS semiLASER and two nWS (MC semiLASER and 2D I-CSE). NAD⁺ was detected with all techniques with a limited visibility for the WS semiLASER. While utilizing nWS concept allows to detect NAD⁺ at the visibility of 66%.

Michal Považan¹, Michael Schär¹, Joseph S Gillen^1,2, and Peter B Barker^1,2
1Department 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

Water suppression employed in ¹H-MR spectroscopy affects signals of exchangeable protons that resonate downfield from water. We have developed a novel method without water suppression for mapping these resonances on a 3T scanner using 2D ¹H-MRSI in combination with binomial spectral spatial excitation and selective refocusing. Acquired data were consistent across all scanned subjects and within the whole FOV, with spectral patterns in agreement with previous single voxel studies.

Jan Willem van der Veen¹ and Jun Shen^1
1Magnetic Resonance Spectroscopy Core, NIH, NIMH, Bethesda, MD, United States

Spectroscopy on high field clinical scanners with limited RF amplitude suffer from large chemical shift artifacts with conventional pulses. Latest developments in adiabatic pulses with modulated gradients like GOIA-WURST pulses offer wide bandwidth at low RF amplitude. The modulated gradient however requires spatial offsets to be added to the phase modulation of the RF pulse. The additional phase offset may cause error due to insufficient digitization of the rapidly changing RF phase. We examined this effect on the voxel profile for two popular pulse parameter sets.

Yihui Huang¹, Jinkui Zhao¹, Zi Wang¹, Di Guo², and Xiaobo Qu^1
1Department of Electronic Science, National Institute for Data Science in Health and Medicine, Xiamen University, Xiamen, China, ²School of Computer and Information Engineering, Xiamen University of Technology, Xiamen, China

Nuclear Magnetic Resonance (NMR) spectroscopy is regarded as an important tool in bio-engineering while often suffers from its time-consuming acquisition. Non-Uniformly Sampling (NUS) method can speed up the acquisition, but the missing FID signals need to be reconstructed with proper method.. In this work, we proposed a deep learning reconstruction method based on unrolling the iterative process of a state-of-the-art model-based low rank Hankel matrix method. Experimental results show that the proposed method provides a better approximation of low rank and preserves the low-intensity signals much better. 

Joon Jang¹, Hyeong Hun Lee¹, Ji-Ae Park², and Hyeonjin Kim^3,4
1Department of Biomedical Sciences, Seoul National University, Seoul, Korea, Republic of, ²Division of Applied RI, Korea Institute of Radiological & Medical Science, Seoul, Korea, Republic of, ³Department of Medical Sciences, Seoul National University, Seoul, Korea, Republic of, ⁴Department of Radiology, Seoul National University Hospital, Seoul, Korea, Republic of

The applicability of generative adversarial networks (GANs) capable of unsupervised anomaly detection (AnoGAN) was investigated in the management of quality of ¹H-MRS human brain spectra. The AnoGAN showed potential in the detection of the spectra with poor SNR or abnormal NAA levels. Despite the fact that those spectra contaminated with ghost, residual water or lipid have never been involved in the training or optimization of the AnoGAN, they were successfully filtered out depending on the intensity of the artifacts. Our unsupervised learning-based approach could be an option in the spectral quality management in addition to the previous supervised learning-based approaches.

Yahang Li^1,2, Zepeng Wang^1,2, and Fan Lam^1,2
1Department of Bioengineering, University of Illinois Urbana-Champaign, Urbana, IL, United States, ²Beckman Institute for Advanced Science and Technology, University of Illinois Urbana-Champaign, Urbana, IL, United States

We report a new method for SNR-enhancing reconstruction of multi-TE MRSI data. Specifically, we designed a deep complex convolutional autoencoder (DCCAE) to learn a nonlinear low-dimensional model of the high-dimensional multi-TE spectra which allowed for effective separation of molecular signals and noise. A constrained reconstruction formulation is used to incorporate the learned model for denoising spatial-temporal reconstruction. The performance of the learned model and the proposed reconstruction method have been evaluated using both simulation and experimental multi-TE $$$^1$$$H-MRSI data. Results obtained demonstrate superior denoising performance achieved by the proposed method over alternative spatial-spectrally constrained denoising strategies.

Kimberly Chan¹ and Anke Henning^1,2
1The University of Texas Southwestern, Dallas, TX, United States, ²Max Planck Institute for Biological Cybernetics, Tübingen, Germany

It has previously been shown that neural networks combined with variable k-space undersampling (MultiNet GRAPPA) is superior to a conventional GRAPPA reconstruction and is feasible at 7T.  Here, MultiNet reconstruction of several new CAIPIRINHA-based variable-density k-space undersampling schemes is investigated.  A new approach to train the neural networks (NN) by augmenting the MRSI data with the non-water suppressed (NWS) data to provide additional self-calibration training data is also introduced and evaluated.  In this study, both are shown here to reduce lipid artifacts and improve metabolic maps at high acceleration factors relative to those previously proposed for MultiNet GRAPPA.

Stanislav Motyka¹, Lukas Hingerl¹, Bernhard Strasser¹, Gilbert Hangel¹, Eva Heckova¹, Asan Agibetov², Georg Dorffner², and Wolfgang Bogner^1
1Department of Biomedical Imaging and Image-guided Therapy, Medical University of Vienna, Wien, Austria, ²Section for Artificial Intelligence and Decision Support (CeMSIIS), Medical University of Vienna, Wien, Austria

A new coil combination method of non-Cartesian kspace MRSI data based on Geometric deep learning is introduced and compared to the conventional image-based coil combination. MRSI data were represented as a graph and a shallow neural network was used to solve the coil combination task. The training data were based on in vivo data and the performance of the network was tested on volunteer data, whose data were never shown to the network. The results were similar to conventional image-domain based coil combination. Thus, a highly accelerated online reconstruction is feasible with this method.

Olivia E Rowe¹, Rangaprakash Deshpande¹, Akila Weerasekera¹, Christopher Stephen^2,3, Robert L Barry^1,4, Florian Eichler^3,5, and Eva-Maria Ratai^1
1Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital and Harvard Medical School, Charlestown, MA, United States, ²Movement Disorders Division and Ataxia Center, Department of Neurology, Massachusetts General Hospital and Harvard Medical School, Boston, MA, United States, ³Center for Rare Neurological Diseases, Department of Neurology, Massachusetts General Hospital and Harvard Medical School, Boston, MA, United States, ⁴Harvard-Massachusetts Institute of Technology Health Sciences & Technology, Cambridge, MA, United States, ⁵Leukodystrophy Clinic, Department of Neurology, Massachusetts General Hospital and Harvard Medical School, Boston, MA, United States

Late-onset GM2-Gangliosidoses (LOGG) are rare lysosomal storage disorders with slowly progressing neurological symptoms that encompass late-onset Tay-Sachs disease (LOTS) and Sandhoff disease (LOSD). We performed sMRI-MRS to discern how cerebellar aberrations may relate to underlying metabolic abnormalities, and their relationships to clinical presentations. Results revealed that structural and metabolic differences between LOTS patients and controls were more prominent than the collective group of LOGG patients vs. controls. Significant clinical associations with imaging-markers also differed when LOTS patients were analyzed separately versus in conjunction with LOSD patients. While LOSD could not be accurately assessed, it appeared to be distinct from LOTS.

Dunja Simicic^1,2,3, Jessie Julie Mosso^1,2,3, Thanh Phong Lê^3,4, Ruud B. van Heeswijk⁵, Ileana Ozana Jelescu^1,2, and Cristina Cudalbu^1,2
1CIBM Center for Biomedical Imaging, Lausanne, Switzerland, ²Animal Imaging and Technology, EPFL, Lausanne, Switzerland, ³Laboratory of Functional and Metabolic Imaging, EPFL, Lausanne, Switzerland, ⁴Geneva School of Health Sciences, HES-SO University of Applied Sciences and Arts Western Switzerland, Geneva, Switzerland, ⁵Department of Radiology, Lausanne University Hospital (CHUV) and University of Lausanne (UNIL), Lausanne, Switzerland

MRSI is a powerful tool for the non-invasive simultaneous mapping of metabolic profiles at multiple spatial positions. This method is highly challenging due to low concentration of metabolites, long measurement times, low SNR, hardware limitations and need for advanced pulse sequences. Denoising based on singular value decomposition has been previously used, but determination of the appropriate thresholds that separate the noise from the signal is problematic leading to possible loss of spatial resolution. Aim of the present study was to implement an improved denoising technique (Marchenko-Pastur principal component analysis) on high resolution MRSI data acquired at 9.4T in the rat-brain.