Synopsis

Free-water elimination allows one to reduce the bias in DTI metrics induced by partial-volume effects. Unfortunately the fitting problem for this model is ill-conditioned. However, it has been recently demonstrated that the introduction of a second dimension determined by the echo-time, leads to a well-conditioned fitting problem. In this work we investigate the experimental design and data analysis pipeline of such experiments in vivo.

Introduction

Methods

The model we adopt here assumes that the diffusion MRI signal originates from two compartments with different transverse relaxation rates and diffusivities, in the slow-exchange limit^1,7 $$S\left( T_\mathrm{E},b ,\mathbf{g}\right)=S_0\left[ f_\mathrm{w}e^{-\frac{T_\mathrm{E}}{T_{2,0}}}e^{-bD_0}+\left(1-f_\mathrm{w}\right)e^{-\frac{T_\mathrm{E}}{T_{2,\mathrm{t}}}}e^{-b\mathbf{g}^\mathrm{T}\mathbf{D}\mathbf{g}}\right]\mathrm{, (1)}$$ where S₀ is the proton density, f_w, T_2,0 and D₀ are the fraction, transverse relaxation time and diffusion coefficient of the free-water compartment and T_2,t and D are the transverse relaxation time diffusion tensor for the tissue compartment. The experimentally controlled parameters are the strength and direction of the diffusion weighting gradient, b and g, and T_E (Fig.1).

Experiments were performed on a healthy volunteer, in a 3T Siemens Trio scanner (Siemens, Erlangen, Germany). Prior written, informed consent was obtained from the volunteer. A Stejskal-Tanner pulse sequence with monopolar pulse field gradients and EPI readout was implemented for this purpose. The sequence was designed to independently control the gradient pulse duration, δ, and separation, Δ, as well as T_E. Thus, the diffusion-dimension and the relaxation-dimension are fully decoupled. Experimental parameters included (Fig. 1): 4 experiments with T_E = 70, 100, 130, 170 ms; b-values = 0 (6 repetitions), 500 (20 directions), 1000 (40 directions) s/mm²; Δ = 34 ms and δ = 19 ms. Additionally, we acquired 8 more echo-times, in the range T_E $\in$ [70, 210] ms. The experiment with T_E = 70 ms was acquired twice with opposite phase-encoding (anterior-posterior, posterior-anterior) in order to correct for EPI distortions. Other parameters were voxel-size = 2³ mm³; matrix-size = 100×100×60; repetition-time, TR = 15 s and GRAPPA acceleration factor = 2.

Eddy current and EPI distortions were corrected using the EDDY toolkit available in FSL.⁸ FWET₂ DTI parameter estimation was performed in three steps:

1. Experimental data with b = 0 s/mm² was used to generate an initial guess for T_2,t using the linear least-squares (LLS) method: $\hat{\gamma}_\mathrm{SE}=\left(W^{\mathrm{T}}_{\mathrm{SE}}W_\mathrm{SE}\right)^{-1}W^{\mathrm{T}}_{\mathrm{SE}}y$, where $W_{\mathrm{SE}}=\left[1,-T_\mathrm{E}\right]^\mathrm{T}$, $\gamma_{\mathrm{SE}}=\left[\ln{S^\mathrm{SE}_0},1/T_{2,\mathrm{t}}\right]^\mathrm{T}$ and y is the measured data. Similarly the initial guess for D was generated using the DWI experiment with T_E = 70 ms using LLS: $\hat{\gamma}_\mathrm{DTI}=\left(W^\mathrm{T}_\mathrm{DTI}W_\mathrm{DTI}\right)^{-1}W^{\mathrm{T}}_{\mathrm{DTI}}y$, where $W_{\mathrm{DTI}}=\left[1,-bg^2_x,-2bg_xg_y,-2bg_xg_z,-bg^2_y,-2bg_yg_z,-bg^2_z,\right]^\mathrm{T}$ and $\gamma_{\mathrm{DTI}}=\left[\ln{S^\mathrm{DTI}_0},D_{11},D_{12}...,D_{33}\right]^\mathrm{T}$.
2. Having had the initial guess for D and T_2,t, an initial guess for f_w was generated using the method proposed by Hoy et al.⁵ and Neto Henriques et al.⁶ using a non-linear least-squares (NLLS) method.
3. Finally FWET₂ was fitted via NLLS minimisation using the initial guess generated in the former steps, with the help of the Levenberg-Marquardt algorithm with in-house Matlab scripts. The power images method⁹ was used to reduce the bias due to background noise prior fitting. Values for D₀ = 3×10^-3 mm²/s and T_2,0 = 500 ms were fixed according to literature values.¹⁰ Thus, the model parameters are $\gamma_\mathrm{FWET2}=\left[S_0,f_\mathrm{w},D_{11},D_{12},...,D_{33},T_{2,\mathrm{t}}\right]$. The NLLS estimator is given by $$\gamma_\mathrm{FWET2}=\arg \min \sum_{i} \left[y_i-S\left(T_{\mathrm{E},i},b_i,\mathbf{g}_i\right)\right]^2\mathrm{. (2)}$$

Results and Discussions

Conclusions

Acknowledgements

References

[1] Pasternak O, et al. Free Water Elimination and Mapping from Diffusion MRI. Magn Reson Med 2009;62:717-730.

[2] Pasternak O, et al. Estimation of Extracellular Volume from Regularized Multi-Shell Diffusion MRI. Med Image Comput Comput Assist Interv 2012;15:305-312.

[3] Ofori E, et al. Increased free-water in the substantia nigra of Parkinson’s disease: a single-site and multi-site study. Neurobiol Aging 2015;36(2):1097-1104.

[4] Hoy AR, et al. Microstructural white matter alterations in preclinical Alzheimer's disease detected using free water elimination diffusion tensor imaging. PLoS One 2017;14;12(3):e0173982.

[5] Hoy AR, et al. Optimization of a free water elimination two-compartment model for diffusion tensor imaging. Neuroimage 2014;103:323-333.

[6] Neto Henriques R, et al. Exploring the potentials and limitations of improved free-water elimination DTI techniques. Proc Intl Soc Magn Reson Med 2017;1787.

[7] Collier Q, et al. Solving the free water elimination estimation problem by incorporating T2 relaxation properties. Proc Intl Soc Magn Reson Med 2017;1783.

[8] Jesper L, et al. An integrated approach to correction for off-resonance effects and subject movement in diffusion MR imaging. NeuroImage 2016;125:1063-1078.

[9] Miller AJ, Joseph PM. The use of power images to perform quantitative analysis on low SNR MR images. Magn Reson Imaging 1993;11(7):1051-6.

[10] Piechniket SK, Functional Changes in CSF Volume Estimated Using Measurement of Water T2 Relaxation. Magn Reson Med 2009;61:579-586.

[11] Gras V, et al. Diffusion-Weighted DESS Protocol Optimization for Simultaneous Mapping of the Mean Diffusivity, Proton Density and Relaxation Times at 3 Tesla. Magn Reson Med 2017;78:130-141.