Henrik Lundell1, Markus Nilsson2, Filip Szczepankiewicz3,4,5, Carl-Fredrik Westin3,6, Daniel Topgaard7, and Samo Lasič8
1Danish Research Centre for Magnetic Resonance, Copenhagen University Hospital Hvidovre, Hvidovre, Denmark, 2Clinical Sciences, Department of Radiology, Lund University, Lund, Denmark, 3Brigham and Women’s Hospital, Boston, MA, United States, 4Harvard Medical School, Boston, Denmark, 5Department of Radiology, Lund University, Lund, Denmark, 6Harvard Medical School, Boston, MA, United States, 7Division of Physical Chemistry, Department of Chemistry, Lund University, Lund, Sweden, 8Random Walk Imaging AB, Lund, Sweden
Synopsis
Multidimensional diffusion encoding isotropic b-tensors eliminate
rotational variance from anisotropic Gaussian diffusion, an essential feature
for estimating microscopic anisotropy. In anisotropic substrates which exhibit
non-Gaussian diffusion on the time scale of the encoding pulse, the variance in
temporal characteristics of diffusion encoding across different directions –
spectral anisotropy (SA) - may introduce a directional variance in apparent
diffusivities. We propose an alternative isotropic encoding with drastically lower
SA, which in turn allows accessing intrinsic signatures of non-Gaussian
diffusion.
Introduction
Multidimensional
diffusion encoding (MDE) utilizes a range of different gradient waveforms and
b-tensor shapes in order to accurately characterize heterogeneous tissues [1–4].
By considering the spectral content of MDE waveforms,
we can probe time-dependent diffusion and tune
its sensitivity to varying compartment sizes [5]. In waveforms that yield isotropic b-tensors,
the spectral content may vary across different directions, which introduces a
rotational variance in substrates with time-dependent anisotropic diffusion [6,7].
We refer to this variation as spectral
anisotropy (SA), which provides a handle for probing correlations between
time dependence and shape of restrictions in heterogeneous samples [8].
Reducing
SA is attractive for clean isotropic measurements. A homogeneous spectral
content can be fairly well realized over rotations in a 2D plane [8,9],
but the 3D case is less straightforward. For instance, the magic angle spinning
of the q-vector (qMAS) has high encoding power at relatively low frequencies along the aperture axis of the trajectory compared to its perpendicular plane. In this work
we propose an alternative isotropic 3D encoding using two subsequent orthogonal
encoding 2D planes and compare its rotational variance to an optimized qMAS
trajectory in a Monte Carlo simulation. Methods
Two encoding waveforms were considered and analyzed in terms of the power spectra |F(ω)|2 of their
corresponding dephasing trajectories [10]:
(1) the original numerically optimized qMAS trajectory [11]
and (2) a new alternative scheme based on two orthogonal planar encoding that we
call 2DORTHO. A planar encoding was extracted from a projection of the qMAS
trajectory with |F(0)|2=0
for all in plane rotations. The full 2DORTHO was realized by repeating the
encoding in the xz- and yz-planes around a refocusing pulse. The z-axis
gradient strength was scaled to provide an isotropic b-tensor.
We
performed Monte Carlo simulations for a 1D system with evenly spaced permeable
membranes [7,12].
The two waveforms were applied with total encoding time of 45 ms including a
refocusing period of 5 ms with 64 uniformly distributed rotations. Membrane
permeability was varied in the range p = 0-0.1 µm/ms, membrane spacing was set
to a = 1-10 µm and the free diffusivity D0 = 2 µm2/ms. Apparent
diffusion coefficients (ADC) and Kurtosis (K) were calculated from the spin
phase distributions for: 1) the simulations in individual directions reflecting
the intrinsic diffusion for different substrate orientations (ADCi,
Ki) and 2) the powder averaged random walks over all directions
reflecting the measured signal for a disordered ensemble (ADCPA, KPA) [7,13].
The latter reflect both the distribution of non-Gaussian intrinsic kurtosis (Ki)
and the contribution from the multi-Gaussian distribution of ADCi
which was calculated as 3) Kv=3·var(ADCi)/⟨ADCi⟩2 [13].
Results and discussion
The
proposed 2DORTHO encoding provides a lower degree of spectral anisotropy in the encoding power
spectrum compared to the originally proposed qMAS as shown in figure 1. The
latter has more encoding power at low frequencies along the z-axis (green
spectrum in figure 1 A). In contrast, the encoding power is more evenly
distributed across the orthogonal axes in 2DORTHO. This is true for any
rotation as shown in high and low frequency portions of the respective power
spectra in figure 1 C) and D). As expected, the lower degree of SA yields a more
narrow distribution of ADC
i in simulations with different rotations
both in the case of impermeable and permeable membranes (top histograms in figure
2 A and B). This is also seen in the spread of the signals shown in
figure 3. The effect of different SA is reflected also in the kurtosis estimates
(lower rows in figure 2), where the lower SA yields smaller K
v. In
case of SA~0, K
PA for the powder averaged signal is probing K
i
without the contribution of K
v. In all cases, we note that the deviations
from mono-exponential signal attenuation due to intrinsic kurtosis are in the
order of 1% at the highest b-value (highlighted regions in figure 3 B and D).
Conclusion
We
propose the alternative isotropic b-tensor encoding 2DORTHO with reduced
spectral anisotropy. This encoding provides rotationally invariant isotropic
encoding even in anisotropic substrates were non-Gaussian effects are
pronounced.Acknowledgements
This project has received funding from the European Research Council (ERC) under the European Union’sHorizon 2020 research and innovation programme (grant agreement No 804746).
References
[1] D. Topgaard, J. Magn. Reson. 275, 98 (2016).
[2] F.
Szczepankiewicz, S. Lasič, D. van Westen, P. C. Sundgren, E. Englund, C. F.
Westin, F. Ståhlberg, J. Lätt, D. Topgaard, and M. Nilsson, Neuroimage 104,
241 (2015).
[3] C.-F.
Westin, F. Szczepankiewicz, O. Pasternak, E. Özarslan, D. Topgaard, H. Knudsen,
and M. Nilsson, MICCAI 14, 209 (2014).
[4] N.
Shemesh, S. N. Jespersen, D. C. Alexander, Y. Cohen, I. Drobnjak, T. B. Dyrby,
J. Finsterbusch, M. A. Koch, T. Kuder, F. Laun, M. Lawrenz, H. Lundell, P. P.
Mitra, M. Nilsson, E. Özarslan, D. Topgaard, and C. F. Westin, Magn. Reson.
Med. 75, 82 (2016).
[5] H.
Lundell, M. Nilsson, T. B. Dyrby, G. J. M. Parker, H. Cristinacce, F. Zhou, D.
Topgaard, and S. Lasic, in In Proc. ISMRM (2017), p. 1086.
[6] T. M.
de Swiet and P. P. Mitra, J. Magn. Reson. - Ser. B 111, 15 (1996).
[7] S. N.
Jespersen, J. L. Olesen, A. Ianu, and N. Shemesh, ArXiv 1 (2017).
[8] H.
Lundell, M. Nilsson, C.-F. Westin, D. Topgaard, and S. Lasic, in In Proc.
ISMRM (2018), p. 887.
[9] H.
Lundell, C. K. Sønderby, and T. B. Dyrby, Magn. Reson. Med. 73, 1171
(2015).
[10] J.
Stepišnik, Phys. B 183, 343 (1993).
[11] D.
Topgaard, Microporous Mesoporous Mater. 178, 60 (2013).
[12] J. E.
Tanner, J Chem Phys 69, 1748 (1978).
[13] J. H.
Jensen, J. A. Helpern, A. Ramani, H. Lu, and K. Kaczynski, Magn. Reson. Med. 53,
1432 (2005).