Synopsis
Multidimensional
diffusion weighting, such as magic angle spinning of the q-vector (q-MAS),
relies on the assumption of multiple Gaussian compartments (MGC). Then the
kurtosis measured with q-MAS can be fully ascribed to ensemble variance of
isotropic diffusivity. However, in compartments with nongaussian diffusion,
anisotropic time dependence of the diffusion tensor imparts orientation
dependence on the q-MAS measured mean diffusivity, which in the presence of
orientation dispersion leads to additional contributions to kurtosis. Yet
another contribution arises from intracompartmental kurtosis. Using simulations
and experiments we demonstrate that q-MAS derived diffusion kurtosis conflates variance in isotropic
diffusivity with dispersion and intracompartmental kurtosis.
Introduction
Currently, generalizations of the classical Stejskal-Tanner
sequence1 for diffusion weighting are being investigated
as a means to characterize microstructure or constrain modeling2-4. Multidimensional
diffusion weighting4,5, which employs
prolonged gradient trajectories, has been proposed to separate diffusion
heterogeneity into sources stemming from orientational and isotropic diffusion
variance5-9. This ability
crucially depends on the assumption of multiple Gaussian compartments (MGC),
but the consequences of violating MGC have not been considered in detail10.
Here we show using analytical calculations, simulations, and experiments, that intracompartmental
diffusion time dependence and kurtosis confound the interpretation of an
isotropic diffusion weighting sequence,
magic angle spinning of the q-vector (q-MAS)11,12.Methods
For the
simulations, we take intrinsic diffusivity $$${{D}_{0}}=1$$$µm2/ms and use the exact solution in a
rectangular box, and Monte Carlo simulations ($$${{10}^{7}}$$$particles) for
regularly spaced barriers with permeability $$$\kappa =0.02$$$µm/ms in 1D, to compute the
signal from the q-MAS waveform in 12. Animal
experiments were preapproved by the local ethics committee and complied with
local and EU laws. Fixed cervical rat spinal cord was imaged on a 16.4T Bruker
Aeon Ascend magnet
with gradients capable of up to 3000 mT/m. A , where a spin-echo
EPI sequence was modified for q-MAS12
with a duration of τ=15
ms and separation of 1.15 ms. Imaging parameters were: bandwidth 500kHz,
2-shots, matrix size 90x54 with a partial Fourier factor of 1.2, in-plane
resolution 75µm by 75µm, slice thickness 1.8mm, 4 averages, and TE/TR
40.5ms/6s. The sequence was repeated for 15 b-values from 0 to 1.5ms/µm2
and 12 different orientations. Additionally, time-dependent intra-and
extra-axonal diffusivities in fixed porcine spinal cord measured with a
Stejskal-Tanner sequence were obtained from 13 (‘+’ branch). Results
Under the MGC
assumption, the diffusion kurtosis $$${{K}_{I}}$$$ associated with q-MAS11
can be related7 to the corresponding
mean (or isotropic) diffusivity $$${{D}_{I}}$$$
and the variance $$${{V}_{I}}$$$ of isotropic diffusivities, $$${{K}_{I}}=3{{V}_{I}}/D_{I}^2$$$.
However, in compartments with nongaussian diffusion, anisotropic time-dependence
of the diffusion tensor imparts orientation dependence on the isotropic diffusivity
measured with q-MAS, which in the presence of orientation dispersion leads to additional
contributions to the variance. Further, intracompartmental diffusion kurtosis
is another source of kurtosis beyond isotropic diffusion variance. These
additional contributions $$$\delta K_{I}^{(1)}$$$ and $$$\delta K_{I}^{(2)}$$$ are:
(i) For identical
axially symmetric microdomains with orientations $$$\mathbf{\hat{u}}$$$ and diffusivities
$$${{D}_{\parallel }}(t)$$$ and $$${{D}_{\bot }}(t)$$$, apparent
isotropic diffusivity is
$${{\tilde{D}}_{\text{I}}}={{\delta
}_{ij}}B_{ij}^{\bot }+{{\langle {{u}_{i}}{{u}_{j}}\rangle
}_{{\mathbf{\hat{u}}}}}B_{ij}^{\Delta }=\text{Tr}({{\text{B}}^{\bot }})+{{\langle
{{\mathbf{\hat{u}}}^{\text{T}}}{{\text{B}}^{\Delta }}\mathbf{\hat{u}}\rangle
}_{{\mathbf{\hat{u}}}}},$$
where $$${{\langle
\cdot \rangle }_{{\mathbf{\hat{u}}}}}$$$ denotes an average over the
orientations $$$\mathbf{\hat{u}}$$$ of the microdomains,
$$B_{ij}^{\bot }=({{\gamma
}^2}/2b)\int_{0}^{T}{d}{{t}_{1}}\int_{0}^{T}{d}{{t}_2}{{G}_{i}}({{t}_{1}}){{G}_{j}}({{t}_2}){{D}_{\bot
}}(|{{t}_{1}}-{{t}_2}|))|{{t}_{1}}-{{t}_2}|$$
$$B_{ij}^{\Delta }=({{\gamma
}^2}/2b)\int_{0}^{T}{d}{{t}_{1}}\int_{0}^{T}{d}{{t}_2}{{G}_{i}}({{t}_{1}}){{G}_{j}}({{t}_2})\left(
{{D}_{\parallel }}(|{{t}_{1}}-{{t}_2}|)-{{D}_{\bot }}(|{{t}_{1}}-{{t}_2}|)
\right)|{{t}_{1}}-{{t}_2}|,$$ and $$$\mathbf{G}$$$ is the diffusion gradient with associated b-value $$$b$$$.
(ii) With an
intracompartmental diffusivity $$$\tilde{D}_{I}^{p}$$$ and kurtosis $$$K_{I}^{p}$$$ in pore $$$p$$$, this additional contribution
becomes
$$\delta K_{I}^{(2)}={{\langle
{{(\tilde{D}_{I}^{p})}^2}K_{I}^{p}\rangle }_{p}}/\tilde{D}_{I}^2$$
where $$${{\langle
\cdot \rangle }_{p}}$$$ denotes an average over the pore population. Thus, q-MAS
derived diffusion kurtosis conflates variance in isotropic diffusivity with
dispersion and intracompartmental kurtosis.
Figure 1 illustrates
that $$${{\tilde{D}}_{\text{I}}}$$$ depends on the orientation of a rectangular
box as function of the side length $$$a$$$, with a variability of up to varying up to more than 20% of the mean. The
corresponding maximum contribution $$$\delta K_{I}^{(1)}$$$ (blue) is
negligible for $$$a<5$$$µm but reaches 0.04 for larger pores. A powder
average limits $$$\delta K_{I}^{(1)}<0.012$$$. Figure 2 shows $$$\delta
K_{I}^{(1)}$$$ for other dimensions, and the red lines indicate regions for which
where $$$\delta
K_{I}^{(1)}$$$ is less than 0.01 (solid) and 0.05 (dashed). Figure 3
illustrates directional variability in pig spinal cord and compares to
$$${{\tilde{D}}_{\text{I}}}$$$ computed
from $$${{D}_{\parallel }}(t)$$$ and $$$D_\bot(t)$$$ obtained in 13. Figure 4
evaluates the directional dependence of $$${{D}_{I}}$$$ measured in rat spinal cord, and estimates $$$\delta K_{I}^{(1)}$$$.
Finally, figure 5 illustrates the importance of intracompartmental kurtosis $$$\delta
K_{I}^{(2)}$$$ for the 4 model systems.
Discussion
Interpretation
of multidimensional diffusion sequences hinges on the MGC assumption. We
demonstrated that violation of MGC prevents separation of orientational and
isotropic heterogeneity with q-MAS, as anisotropic time-dependence and
intracompartmental kurtosis led to sizable relative contributions to $$${{K}_{I}}$$$:
for comparison, reported values5,8 are $$${{K}_{I}}\approx 0.3-0.6$$$. This was experimentally confirmed
by a significant dependence on gradient frame orientation for q-MAS in fixed
spinal cord. We note that double diffusion encoding2,3 also provides
the population variance of diffusion tensors without the MGC assumption14,15,
and that other multidimensional diffusion sequences that are less sensitive
with less
sensitivity to nongaussianity could be designed.Conclusion
Experimentally
observed deviations from the MGC assumption call for caution when interpreting
multidimensional diffusion weighting. Specifically, additional important contributions to
variance in diffusivity arise from dispersion of compartments with anisotropic
diffusion time dependence, and intracompartmental diffusion kurtosis.Acknowledgements
SNJ is supported by the Danish National Research
Foundation (CFIN), and The Danish Ministry of Science, Innovation, and
Education (MINDLab). Funding from the
European Research Council (ERC) under the European Union’s Horizon 2020
research and innovation programme (Starting Grant, agreement No. 679058)
supports NS, and funding from EPSRC grant number M507970 supports AI. We thank
Dr. Daniel Nunes for tissue extraction.References
1. Stejskal, E.O.
and J.E. Tanner, Spin Diffusion
Measurements: Spin Echoes in the Presence of a Time-Dependent Field Gradient.
The Journal of Chemical Physics, 1965. 42(1):
p. 288-292.
2. Cheng, Y. and
D.G. Cory, Multiple scattering by NMR.
J. Am. Chem. Soc., 1999. 121(34): p.
7935-7936.
3. Shemesh, N., et
al., Conventions and nomenclature for
double diffusion encoding NMR and MRI. Magn Reson Med, 2016. 75(1): p. 82-7.
4. Topgaard, D., Multidimensional diffusion MRI. J Magn
Reson, 2017. 275: p. 98-113.
5. Westin, C.F., et
al., Q-space trajectory imaging for
multidimensional diffusion MRI of the human brain. Neuroimage, 2016. 135: p. 345-62.
6. Lampinen, B., et
al., Neurite density imaging versus
imaging of microscopic anisotropy in diffusion MRI: A model comparison using spherical
tensor encoding. Neuroimage, 2017. 147:
p. 517-531.
7. Lasic, S., et
al., Microanisotropy imaging:
quantification of microscopic diffusion anisotropy and orientational order
parameter by diffusion MRI with magic-angle spinning of the q-vector. Frontiers
in Physics, 2014. 2.
8. Szczepankiewicz,
F., et al., The link between diffusion
MRI and tumor heterogeneity: Mapping cell eccentricity and density by
diffusional variance decomposition (DIVIDE). Neuroimage, 2016. 142: p. 522-532.
9. Szczepankiewicz,
F., et al., Quantification of microscopic
diffusion anisotropy disentangles effects of orientation dispersion from
microstructure: applications in healthy volunteers and in brain tumors.
Neuroimage, 2015. 104: p. 241-52.
10. Ianus, A., et al. Metrics of microscopic anisotropy: a
comparison study. in Proc. Int. Soc.
Magn. Reson. Med. 2016. Singapore, Singapore.
11. Eriksson, S., S.
Lasic, and D. Topgaard, Isotropic
diffusion weighting in PGSE NMR by magic-angle spinning of the q-vector. J
Magn Reson, 2013. 226: p. 13-8.
12. Nilsson, M., et
al., Liquid crystal phantom for
validation of microscopic diffusion anisotropy measurements on clinical MRI
systems. Magn Reson Med, 2017.
13. Jespersen, S.N.,
et al., Diffusion time dependence of
microstructural parameters in fixed spinal cord. Neuroimage, 2017.
14. Jespersen, S.N.,
et al., Orientationally invariant metrics
of apparent compartment eccentricity from double pulsed field gradient
diffusion experiments. NMR Biomed, 2013. 26(12): p. 1647-62.
15. Jespersen,
S.N., Equivalence of double and single
wave vector diffusion contrast at low diffusion weighting. NMR Biomed,
2012. 25(6): p. 813-8.