Quantitative Magnetization Transfer Imaging Made Easy with qMTLab: Software for Data Simulation, Analysis and Visualisation
Jean-François Cabana1, Ye Gu2, Mathieu Boudreau3, Ives R. Levesque3, Yaaseen Atchia4, John G. Sled4, Sridar Narayanan3, Douglas L. Arnold3, Bruce G. Pike5, Julien Cohen-Adad6, Tanguy Duval6, Manh-Tung Vuong6, and Nikola Stikov6

1Medical Physics, University of Montreal, Montreal, QC, Canada, 2NeuroRX, Montreal, QC, Canada, 3McGill University, Montreal, QC, Canada, 4University of Toronto, Toronto, ON, Canada, 5University of Calgary, Calgary, AB, Canada, 6Ecole Polytechnique, Montreal, QC, Canada

Synopsis

We have developed a free, open source software (qMTLab) that unifies the most widely used quantitative magnetization transfer imaging (qMTI) methods in a simple and easy to use graphical interface. qMTLab allows to easily simulate qMTI data, compare the performance of the methods under various experimental conditions, define new acquisition protocols, fit acquired data, and visualize the fitted parameter maps. In this presentation, we will offer a brief introduction on the theory behind qMTI, present the current acquisition and analytical methods, and present the functionality of the qMTLab software and its utility in basic and clinical research.

Purpose

Quantitative magnetization transfer imaging (qMTI) increases specificity to macromolecular content in tissue by modeling the exchange process between the liquid and the macromolecular pool. However, its use has been limited, in part due to the need to write complicated in-house software for modeling and data analysis. We have developed a free open source software, qMTLab (https://github.com/neuropoly/qMTLab/), that unifies the most widely used qMTI methods and allows one to easily simulate qMTI data, compare the performance of the methods under various experimental conditions, define new acquisition protocols, fit acquired data, and visualize the fitted parameters maps. By providing free software with a simple and easy to use graphical interface, we hope to make qMTI accessible to a greater number of investigators and facilitate the development and optimization of protocols.

Outline of Content

We will begin by presenting an introduction on the theory behind qMTI. We will then explain the current acquisition methods and the analytical solutions that have been developed to extract useful physical parameters about this model. Finally, we will present the functionality of the qMTLab software and demonstrate its utility.

Methods

Theory

In vivo qMTI methods are based on a model that describes hydrogen nuclei in tissues as composed of two distinct pools: a free pool of highly mobile protons associated with water, and a restricted pool of less mobile protons residing in macromolecules (Figure 1)1,2. This model is characterised by seven parameters: pool size ratio (F); exchange rate (either kr or kf, the two being related by F = kf/kr); longitudinal relaxation rates (R1f, R1r); transverse relaxation times (T2f, T2r) and the equilibrium magnetization (M0f, with M0r = F×M0f). Under certain assumptions, analytical solutions can be derived and a signal equation can be fitted to the data, from which the qMTI parameters are derived.

Three acquisition protocols can be used for qMTI. MT-SPGR is a steady-state acquisition which combines off-resonance pulsed saturation (variable duration and power) of the restricted pool with an SPGR readout (Figure 2a)3,4. SIR-FSE uses a selective inversion recovery (SIR) sequence with variable inversion times followed by a fast spin echo (FSE) readout (Figure 2b)5,6. In MT-bSSFP, a steady-state acquisition is performed by applying a series of equally spaced, on-resonance RF pulses of variable duration and flip angles (Figure 2c)7.

qMTLab software presentation

The software consists of two parts: 1) a qMTI data simulator, and 2) a qMTI data fitting and visualization interface. The simulation part allows users to generate synthetic qMTI data using the above described protocols, evaluate how well they perform under known ground-truth parameters, determine the most appropriate acquisition parameters, and evaluate how fitting constraints impact the results. The data fitting part provides an interface to import qMTI data, fit models to the data and visualize the resulting parameters maps. An example of the graphical user interface is presented in Figure 5.

Synthetic data

Three modes of simulation are offered (Figure 6 a-c): 1) Single Voxel Curve, to simulate MT data from a single voxel; 2) Sensitivity Analysis, allows systematic variation of one MT parameter, over a defined range and number of points, while keeping the others fixed; and 3) Multi Voxel Distribution, where any parameter combination is allowed to be varied simultaneously for a number of voxels.

Data fitting

The data fitting part of qMTLab (Figure 6d) provides a simple interface to load acquired qMTI data, fit a model and visualize the resulting parameters maps. The user can provide a qMTI data file, select or define the protocol used for acquisition and set the fitting options. The user can also load a mask to constrain the areas to be analyzed, an observed R1 map for R1f evaluation, a B1 map for MT pulse power or excitation flip angle correction and a B0 map for correction of offset frequencies. Fitting results are displayed as parametric maps with user-controllable color scales.

Conclusion

The ability to map quantitative values such as the rate of magnetization exchange between free and restricted protons or the ratio of the proton pools makes qMTI a valuable tool in characterizing tissue microstructure. By providing free software that gives end users a simple and easy-to-use graphical interface, we hope to make qMTI accessible to a greater number of investigators and facilitate the development and optimization of protocols for research in clinical populations. The open-source nature of qMTLab makes it particularly useful, in that it is subject to continuous improvement and that users can customize it and add functionality to fit individual needs.

The qMTLab software can be downloaded for free at: https://github.com/neuropoly/qMTLab/

Acknowledgements

No acknowledgement found.

References

1. Henkelman RM, Huang X, Xiang QS, Stanisz GJ, Swanson SD, Bronskill MJ. 1993 Quantitative interpretation of magnetization transfer. Magn Reson Med;29(6):759-66.

2. Henkelman RM, Stanisz GJ, Graham SJ. 2001 Magnetization transfer in MRI: a review. NMR Biomed;14(2):57-64.

3. Pike GB. 1996 Pulsed magnetization transfer contrast in gradient echo imaging: A two-pool analytic description of signal response. Magnetic Resonance in Medicine;36(1):95-103.

4. Sled JG, Pike GB. 2000 Quantitative interpretation of magnetization transfer in spoiled gradient echo MRI sequences. J Magn Reson;145(1):24-36.

5. Gochberg DF, Gore JC. 2003 Quantitative imaging of magnetization transfer using an inversion recovery sequence. Magn Reson Med;49(3):501-5.

6. Gochberg DF, Gore JC. 2007 Quantitative magnetization transfer imaging via selective inversion recovery with short repetition times. Magn Reson Med;57(2):437-41.

7. Bieri O, Scheffler K. 2007 Optimized balanced steady-state free precession magnetization transfer imaging. Magn Reson Med;58(3):511-8.

Figures

Two-pool model of magnetization transfer exchange in the absence of applied RF. Each pool is divided into two groups, representing the state of magnetization (adapted from Henkelman1).

qMTI acquisition sequences. a: MT-SPGR sequence. Timings are: tmt = MT pulse; ts = Spoiling gradients; tp = Excitation pulse; tr = Recovery. b: SIR-FSE sequence. Timings are: td = Delay; ti = Inversion recovery. c: MT-bSSFP sequence. TRF = RF pulse duration. TR = Repetition time.

qMTLab main interface, showing the SIR-FSE fitted curve for a single voxel. Left panel is the menu where the user can choose the method and the task to perform; center panel is where results are displayed; right window is the options panel, where all the parameters are set.

The different task panels in qMTlab. a Single Voxel Curve; b Sensitivity Analysis; c Multi Voxel Distribution; d qMTI Data Fit



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
3205