Osprey: Open-Source Processing, Reconstruction & Estimation of Magnetic Resonance Spectroscopy Data
Poster No:
2124
Submission Type:
Abstract Submission
Authors:
Georg Oeltzschner1,2, Helge Zöllner1,2, Richard Edden1,2
Institutions:
1Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University, Baltimore, MD, 2F. M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute, Baltimore, MD
First Author:
Georg Oeltzschner
Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University|F. M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute
Baltimore, MD|Baltimore, MD
Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University|F. M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute
Baltimore, MD|Baltimore, MD
Co-Author(s):
Helge Zöllner
Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University|F. M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute
Baltimore, MD|Baltimore, MD
Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University|F. M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute
Baltimore, MD|Baltimore, MD
Richard Edden
Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University|F. M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute
Baltimore, MD|Baltimore, MD
Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University|F. M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute
Baltimore, MD|Baltimore, MD
Introduction:
Modern magnetic resonance spectroscopy (MRS) data analysis requires elaborate preprocessing, interfacing with external fitting software, and tissue and relaxation corrections of the results obtained with the external software. Well-resourced labs frequently rely on in-house code for such tasks, but a widely used standardized pipeline is not available. Additionally, the default linear-combination modeling software is an expensive, closed-source, commercial product with limited on-going development. As a result, the entry threshold for new labs looking to apply MRS is high, the methods applied are heterogeneous and often poorly described in the literature, and the future is uncertain.
Here we describe a new MATLAB-based toolbox "Osprey" which streamlines all steps of state-of-the-art pre-processing, linear-combination modeling, tissue correction, quantification, and visualization of MRS data into a single environment. The Osprey framework is designed in a modular way to flexibly adopt new methods and encourage community contribution.
Here we describe a new MATLAB-based toolbox "Osprey" which streamlines all steps of state-of-the-art pre-processing, linear-combination modeling, tissue correction, quantification, and visualization of MRS data into a single environment. The Osprey framework is designed in a modular way to flexibly adopt new methods and encourage community contribution.
Methods:
The Osprey workflow is based on FID-A functions (Simpson 2017) and consists of seven modules: JOB, LOAD, PROCESS, FIT, COREG, SEG, and QUANTIFY. After defining a JOB by specifying spectroscopic data and structural images, LOAD stores the raw data in an Matlab data structure. PROCESS performs coil-combination, alignment of individual averages, eddy-current correction, water removal, baseline correction, and others. FIT models the processed spectra in the frequency domain (concatenating the real and imaginary parts) through constrained non-linear least-squares optimization. COREG creates a MRS voxel mask and co-registers it to a structural image. SEG invokes SPM12 segmentation to derive fractional tissue volumes for gray matter, white matter, and CSF. Quantify calculates quantitative outputs: ratios /tCr; CSF-corrected; tissue-and-relaxation-corrected.
A GUI is available (Fig. 1). Osprey supports batch processing, and recognizes most common file formats. Currently, single-voxel conventional and J-difference-edited data are supported. Basis sets and support for other sequences are continuously added.
Osprey further exports pre-processed data in formats readable by LCModel, Tarquin, and jMRUI. Osprey fitting results are exported in CSV format for further analysis in statistical software.
A GUI is available (Fig. 1). Osprey supports batch processing, and recognizes most common file formats. Currently, single-voxel conventional and J-difference-edited data are supported. Basis sets and support for other sequences are continuously added.
Osprey further exports pre-processed data in formats readable by LCModel, Tarquin, and jMRUI. Osprey fitting results are exported in CSV format for further analysis in statistical software.
·Fig. 1: Osprey GUI showing the results of the Process (upper panel) and Fit (lower panel) modules.
Results:
Representative linear-combination modeling results of a PRESS dataset and a GABA-edited MEGA-PRESS dataset are shown in Fig. 2a and 2b.
·Fig. 2: (a) Representative linear-combination modeling of PRESS and (b) MEGA-PRESS data.
Conclusions:
Since MRS is a quantitative technique, the results of MRS experiments depend substantially on the way that data are processed, modelled, and evaluated. While widely used de-facto-standardized processing and analysis toolboxes have been developed for many other quantitative MRI modalities, no such framework currently exists for MRS. As a result, most researchers have developed their own legacy code to prepare their data for third-party quantification software.
This practice is problematic for a number of reasons: a) methodological heterogeneity and opacity diminish comparability and reproducibility of quantitative MRS studies; b) benchmarking and adaptation of methodological progress is considerably slowed down; c) researchers new to the field experience a high-level entry threshold; d) strong dependency on engagement, support, and funding situation of third-party software developers leaves the community vulnerable.
The new toolkit 'Osprey' has been designed to address the lack of a freely available software that unifies all steps of modern MRS data analysis in a common framework. Osprey bundles robust, peer-reviewed data processing methods into a modular workflow that is easily augmented by community developers. In particular, the quantification model is fully accessible, modifiable, and exchangeable, allowing researchers to develop open-source alternatives to third-party fitting software.
The Osprey source code is publicly available at https://github.com/schorschinho/osprey.
This practice is problematic for a number of reasons: a) methodological heterogeneity and opacity diminish comparability and reproducibility of quantitative MRS studies; b) benchmarking and adaptation of methodological progress is considerably slowed down; c) researchers new to the field experience a high-level entry threshold; d) strong dependency on engagement, support, and funding situation of third-party software developers leaves the community vulnerable.
The new toolkit 'Osprey' has been designed to address the lack of a freely available software that unifies all steps of modern MRS data analysis in a common framework. Osprey bundles robust, peer-reviewed data processing methods into a modular workflow that is easily augmented by community developers. In particular, the quantification model is fully accessible, modifiable, and exchangeable, allowing researchers to develop open-source alternatives to third-party fitting software.
The Osprey source code is publicly available at https://github.com/schorschinho/osprey.
Modeling and Analysis Methods:
Methods Development
Neuroinformatics and Data Sharing:
Databasing and Data Sharing
Workflows 2
Novel Imaging Acquisition Methods:
MR Spectroscopy 1
Physiology, Metabolism and Neurotransmission :
Physiology, Metabolism and Neurotransmission Other
Keywords:
Acquisition
Data analysis
GABA
Glutamate
Magnetic Resonance Spectroscopy (MRS)
Modeling
MR SPECTROSCOPY
Neurotransmitter
Workflows
1|2Indicates the priority used for review