This is a list of single-voxel MRS datasets.

2SPECIAL vGRAPPA

This algorithm decomposes simultaneously acquired multi-voxel data to their respective origin regions. Based on the split-slice GRAPPA code by Steen Moeller, University of Minnesota, USA.

2SPECIAL vGRAPPA
Developer Layla Tabea Riemann, M.Sc. (Physikalisch-Technische Bundesanstalt, PTB)
Format SIEMENS DAT
Sequence (2)SPECIAL
License CC BY-NC-SA 4.0
Credit Please cite the following publication if you use the vGRAPPA algorithm: L.T.Riemann, C.S.Aigner, R.Mekle, S.Schmitter, B.Ittermann, A.Fillmer; Fourier-based decomposition approach for simultaneous acquisition of 1H spectra from two voxels in vivo at short echo times, Proc. Int. Soc. Magn. Reson. Med. 2021:007
Contact layla.riemann@ptb.de

Link to Dataset 

Basis sets for PRESS, STEAM and sLASER for multiple vendors

These basis sets were generated using the exact waveforms, timings and 128^3 spatial points to appropriately accommodate the sidebands for each vendor/sequence combination. More details about the acquisition can be found in the reference below.

Basis sets for PRESS, STEAM and sLASER for multiple vendors
Developer Karl Landheer, Kelley Swanberg, Christoph Juchem
Format Both .RAW and .mat (INSPECTOR, readable in MATLAB) are provided. All files were generated using MATLAB
Sequence PRESS, sLASER, STEAM
License BSD3
Credit Please cite the following publication if you use this dataset: Landheer K, Swanberg, K, Juchem C. Magnetic resonance Spectrum simulator (MARSS), a novel software package for fast and computationally efficient basis set simulation. NMR Biomed. 2019, e4219. doi.org/10.1002/nbm.4129
Contact cwj2112@columbia.edu

Link to Dataset Publication

Synthetic brain MR spectroscopy data - sLASER - TE: 35ms - 3T

Synthetic MR spectroscopy data for quantification purposes (included ground-truth concnetration levels) of use for benchmarking (model fitting, machine learning or neural networks algorithms) or of use in deep learning contexts for pre-training, fine-tuning or ablation study. Various dataset size and distribution of concentrations for brain spectra are included (for full description, see README.md file available at the URL reported below or at https://github.com/bellarude/MRS_detasets). Variation of shim quality, SNR levels and MMBG contribution are included. Spectra are free of 0th- and 1st-order phase offset as well as freqeuncy drifts. (optional) Additional downscaled water signal is added (for some datasets) for quantification referencing. (otpional) Spectral preprocessing to spectrogram is included. For further details see: https://doi.org/10.1002/mrm.29561

Synthetic brain MR spectroscopy data - sLASER - TE: 35ms - 3T
Developer Rudy Rizzo, MSc and Roland Kreis, PhD
Format Matlab .dat
Sequence sLASER, TE 35ms, 3T
License CC-BY 4.0 (https://creativecommons.org/licenses/by/4.0/legalcode)
Credit Please cite Rizzo R. et al, Quantification of MR spectra by deep learning in an idealized setting: Investigation of forms of input, network architectures, optimization by ensembles of networks, and training bias, Magn. Reson. Med. 89(5):1707-1727 (2023).
Contact roland.kreis@insel.ch

Link to Dataset Publication

Big GABA

This repository contains a subset of a large-scale multi-vendor, multi-site collection of single-voxel MRS and structural MRI datasets that were acquired internationally across 26 research sites on 3T MRI scanners from the three major vendors (GE, Philips, Siemens). Each site acquired up to 12 datasets.

Big GABA
Developer Richard AE Edden
Format SDAT, TWIX, P
Sequence MEGA-PRESS, PRESS
License open-source
Credit If using these data, please cite: (Mikkelsen M et al. Big GABA: Edited MR spectroscopy at 24 research sites. NeuroImage 2017;159:32–45. doi:10.1016/j.neuroimage.2017.07.021), (Mikkelsen M et al. Big GABA II: Water-referenced edited MR spectroscopy at 25 research sites. NeuroImage 2019;191:537–548. doi:10.1016/j.neuroimage.2019.02.059), and (Povazan M et al. Comparison of multivendor single-voxel MR spectroscopy data acquired in healthy brain at 26 sites. Radiology 2020. doi:10.1148/radiol.2020191037), and acknowledge NIH grant R01 EB016089.
Contact gabamrs@gmail.com

Link to Dataset Publication

Fitting challenge data, MRS Workshop 2016

Synthetic data mimicking single-voxel 3 T PRESS TE = 30 ms MRS spectra. Data were simulated using ideal pulses and the known metabolites concentrations in the human brain. The experimentally acquired macromolecular contribution was also used. The following things were modulated in the datasets: non-Lorentzian broadening, concentrations of GABA, GSH and macromolecules, signal-to-noise (SNR) variations, and artifacts. These data was used in the Fitting Challenge organized as part of the ISMRM Workshop on MR Spectroscopy: From Current Best Practice to Latest Frontiers, Lake Constance, Germany, 2016.

Fitting challenge data, MRS Workshop 2016
Developer Malgorzata Marjanska, Dinesh Deelchand, Roland Kreis
Format text, LCModel (.RAW and .h2o), and jMRUI (ascii text)
Sequence PRESS TE = 30 ms, simulated
License BSD3
Credit Please acknowledge developers as listed above and list following URL: https://www.ismrm.org/workshops/Spectroscopy16/mrs_fitting_challenge/
Contact gosia@umn.edu

Link to Dataset 

MM human brain spectra 9.4T - TE series / Fit Settings LCModel

This are TE series (TE = 24,32,40,52,60 ms) DIR semi-LASER macromolecular spectra. They were measured in the occipital lobe (in GM rich and WM voxels) of the human brain. The provided data are averages from 11 subjects. Additionally, the fitsettings and sample fits are provided. An explanation on how to set up the fit settings file is provided in the Readme.MD.

MM human brain spectra 9.4T - TE series / Fit Settings LCModel
Developer Tamas Borbath
Format LCModel .RAW
Sequence sLASER
License BSD3
Credit Please cite the following publication if you use the MM human brain spectra 9.4T dataset. Murali Manohar S, Borbath T, Wright AM, Soher B, Mekle R, Henning A. T2 relaxation times of macromolecules and metabolites in the human brain at 9.4 T. Magnetic resonance in medicine. 2020;84:542–58.
Contact tamas.borbath@tuebingen.mpg.de

MRSHub Data  Publication

MM Consensus Data Collection

This repository contains a number of pre-processed macromolecule (metabolite-nulled) spectra in Varian format (sometimes converted from Siemens format), acquired at different field strengths from different mammals, including human. These data are represented in Figure 1 of the recently published consensus on macromolecule acquisition and handling.

MM Consensus Data Collection
Developer Ivan Tkáč (4 and 7T, human; 9.4T rat, mouse, cat), Anke Henning (9.4T human), Cristina Cudalbu (14.1T, rat), Dinesh Deelchand (3T, human)
Format Varian or Siemens converted in Varian
Sequence STEAM, SPECIAL, sLASER
License BSD3
Credit Please cite the following publications if you use data from this dataset collection: Tkáč et al., Magn Reson Med 2003; 50: 24-32 (MM 9.4T rat); Tkáč et al., Magn Reson Med 2004; 52: 478-484 (MM 9.4T mouse, cat); Tkáč et al., Magn Reson Med 2009; 62: 868-879 (MM and spectra 4T and 7T); Cudalbu et al., J Alzheimers Dis. 2012;31 Suppl 3:S101-15. doi: 10.3233/JAD-2012-120100 (MM 14.1T rat)
Contact cristina.cudalbu@epfl.ch; ivan@cmrr.umn.edu

MRSHub Data 

Monte Carlo simulations of synthetic data

Monte Carlo simulation to investigate the relationship between CRLB and SD for all fitted parameters from linear combination modeling. More details about the acquisition can be found in the reference below.

Monte Carlo simulations of synthetic data
Developer Karl Landheer, Christoph Juchem
Format .mat file (readable in MATLAB). Synthetic spectra were generated from MARSS, fitting was performed in INSPECTOR. Both MARSS and INSPECTOR are based in MATLAB.
Sequence Synthesized spectra were simulated for a TE = 20.1 ms sLASER sequence
License BSD3
Credit Please cite the following publication if you use this dataset: Landheer K, Juchem C. Are Cramér-Rao Lower Bounds an Accurate Estimate for Standard Deviations in In Vivo Magnetic Resonance Spectroscopy? NMR Biomed. [In press]
Contact cwj2112@columbia.edu

Link to Dataset 

MRSDB

MRSDB: A Scalable Multisite Data Library for Clinical and Machine Learning Applications of Magnetic Resonance Spectroscopy

MRSDB
Developer Sam H. Jiang, BS
Format CSV (from Siemens TWIX, processed in LCModel)
Sequence PRESS
License none
Credit Center for Clinical Spectroscopy, Brigham and Women's Hospital and Harvard Medical School
Contact hjiang@bwh.harvard.edu

Link to Dataset 

Multi-site Frequency Drift MR Spectroscopy

This dataset consists of PRESS data acquired before and after running a heavy gradient duty fMRI sequence using standardized protocols for GE, Philips and Siemens scanners. 1) Pre-fMRI PRESS (TR/TE = 5000/35 ms; FA = 90°; 64 transients; no water suppression; voxel size = 2 × 2 × 2 cm3; second-order shim; scan duration = 5:20 min); 2) BOLD-weighted EPI based on the ADNI-3 protocol (Weiner et al., 2017) (TR/TE = 3000/30 ms; 197 dynamics; EPI factor/echo train length = 31/64, duration = 10 min); -3) Post-fMRI PRESS (same parameters as for pre-fMRI PRESS except 360 transients and duration = 30 min).

Multi-site Frequency Drift MR Spectroscopy
Developer SCN Hui et al (Full author list bit.ly/MRSFrequencyDrift)
Format SDAT, TWIX, P
Sequence PRESS
License This dataset is distributed freely under the Creative Commons Attribution-NonCommercial-ShareAlike license and consensus from co-authors.
Credit Please cite the following publications if you use this dataset: - Hui SCN, Mikkelsen M, Zöllner HJ et al. Frequency Drift in MR Spectroscopy: An 87-scanner 3T Phantom Study, ISMRM 29th Annual Meeting, May 15-20, 2021, Vancouver - Hui SCN, Mikkelsen M, Zöllner HJ, et al. Frequency Drift in MR Spectroscopy at 3T. NeuroImage (under review)
Contact stevehui@jhu.edu

Link to Dataset 

NeuroMET - SPECIAL MRS Reproducibility

Magnetic Resonance Spectroscopy Data acquired in 9 healthy volunteers using a SPECIAL Localization with three different adiabatic inversion pulses (hyperbolic secant, WURST, GOIA) at a 7T Magnetom whole-body system (Siemens Healthineers, Erlangen, Germany). Every volunteer was examined 4 times (twice on day one including a repositioning between the measurements, and twice on day two a week later without repositioning between alike measurements) to investigate the repeatability and reproducibility. Please find more details in L.T. Riemann, C.S. Aigner, S.L.R. Ellison, R. Brühl, R. Mekle, S. Schmitter, O. Speck, G. Rose, B. Ittermann, A. Fillmer, Assessment of Measurement Precision in Single Voxel Spectroscopy at 7 T: Towards Minimal Detectable Changes of Metabolite Concentrations in the Human Brain In-Vivo, Magn Reson Med DOI 10.1002/mrm.29034 (in press).

NeuroMET - SPECIAL MRS Reproducibility
Developer Layla Tabea Riemann, Christoph Stefan Aigner, Stephen L.R. Ellison, Rüdiger Brühl, Ralf Mekle, Sebastian Schmitter, Oliver Speck, Georg Rose, Bernd Ittermann, Ariane Fillmer
Format Siemens .dat, npy, xls
Sequence SPECIAL
License CC BY-NC-SA 4.0
Credit Please cite the publication DOI 10.1002/mrm.29034 as well as the data set DOI 10.5281/zenodo.5500320 if you use the NeuroMET - SPECIAL MRS Reproducibility dataset.
Contact ariane.fillmer@ptb.de

Link to Dataset Publication

fMRS in pain

This is functional MRS (3T) during an in-scanner painful intervention. It contains 15 subjects’ baseline and functional spectra (TE = 22 ms) acquired at 3T from the human anterior cingulate cortex.

fMRS in pain
Developer Jessica Archibald, Erin MacMillan & John Kramer
Format Philips SDAT, data/list, nii
Sequence PRESS
License CC0 Creative Commons
Credit Please cite the following publication if you use the fMRS in pain dataset: Archibald J., MacMillan EL., Graf C., Kozlowski P., Laule C., Kramer J.L.K. ''Metabolite activity in the anterior cingulate cortex during a painful stimulus using functional MRS''. Sci Reports 2020
Contact jessica.archibald@ubc.ca

Link to Dataset Publication

9.4T MR Spectra from rat hippocampus with LCModel quantification and the corresponding basis set

This repository contains the LCModel quantifications of 9.4T spectra acquired in hippocampus from 7 rats. The spectra were quntified using six different DKNTMN (spline stiffness) values (0.1, 0.25, 0.4, 0.5, 1, 5). In the folder Control_files_Basis_set you can find all the control files used in this quantification along with the corresponding basis set (metabolites/simulated using NMRScopeB from jMRUI and in vivo parameters + full MM spectrum).

9.4T MR Spectra from rat hippocampus with LCModel quantification and the corresponding basis set
Developer Dunja Simicic and Cristina Cudalbu (CIBM/EPFL)
Format Varian, .RAW, .mat, .txt, .pdf, .control, .basis
Sequence SPECIAL
License BSD3
Credit Please cite the Zenodo resource linked below if you use the dataset (publication in progress).
Contact dunja.simicic@epfl.ch; cristina.cudalbu@epfl.ch

Link to Dataset