Giorgia Milotta1, Nadège Corbin1,2, Antoine Lutti3, Siawoosh Mohammadi4,5, and Martina Callaghan1
1Wellcome Centre for Human Neuroimaging, UCL Queen Square Institute of Neurology, University College London, London, United Kingdom, 2Centre de Résonance Magnétique des Systèmes Biologiques, UMR5536, CNRS/University Bordeaux, Bordeaux, France, 3Laboratory for Research in Neuroimaging, Department for Clinical Neuroscience, Lausanne University Hospital and University of Lausanne, Lausanne, Switzerland, 4Department of Systems Neurosciences, University Medical Center Hamburg-Eppendorf, Hamburg, Germany, 5Department of Neurophysics, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany
Synopsis
Quantitative
relaxometry in the brain is appealing because of its microstructural
sensitivity. Estimating bulk parameters assumes a single relaxation time per
voxel, which is only valid when the residency time is short with respect to T1.
In reality, the relative contribution of sub-compartments will depend on flip
angle and echo time. Here, we simulate a two-compartment model (myelin and
intra-extracellular water) and estimate T1 with FA-specific signals
derived via three estimation schemes. We quantify the impact of myelin water
fraction, residency time and transmit field inhomogeneity on these estimates
and find good correspondence with in vivo T1 estimates at 7T.
Introduction
Quantitative
relaxometry in the brain is appealing because of its microstructural
sensitivity. However, the estimation of bulk parameters assumes a single
relaxation time per voxel. This is only valid under the assumption of fast
exchange, whereby the residency time in a given microenvironment is short with
respect to the relaxation time. In reality sub-compartments contribute to the
measured signal and vary with flip angle (FA) and echo time, TE (1,2).
Here
we simulate two exchanging water pools and derive FA-specific signals via three
estimation schemes before estimating a bulk T1 assuming a single
compartment. The sensitivity of these estimates to myelin water fraction (fMW),
residency time and transmit field inhomogeneity (B1+eff
= actual FA/nominal FA) is quantified, and compared to in vivo T1 estimates at 7T.Methods
Simulations
The
spoiled gradient echo signal for a two-compartment model was simulated using
the Bloch-McConnell equations implemented with the EPG formalism (3). Spoiling assumed a 6π
moment per TR and an RF spoiling increment of 144°. B
1+eff
was varied according to [0.4:0.2:1.6]. Multi-echo signals were simulated with
FA of 6° and 26° respectively, with TR=19.5ms and TE=[2.56:2.30:14.5]ms. T
1
was estimated analytically (4) using signals S
1 and S
2
derived from these data in three different ways (Figure 1):
- Single-echo: FA-specific signals at
TE=4.86ms.
-
Per-contrast: intercepts estimated
from a log-linear fit for each FA separately.
-
ESTATICS: pooling both FA and
performing a single log-linear fit assuming a common decay. This method is appealing because it exhibits
enhanced robustness to low SNR and motion (2).
Simulations
were carried out with 7T-relevant T
1 and T
2 times for
myelin (T
1MW = 280ms, T
2MW=T
2*
MW =
8ms for simplicity) and intra-extracellular (T
1IE = 1700ms, T
2IE=T
2*
IE
= 36ms) water. f
MW spanned a range, [2:2:20]%, expected to
encompass white mater (WM) and grey matter (GM) (5). A directional (myelin to intra-extracellular)
residency time was varied, [100:100:500]ms, to investigate different exchange
regimes (1). Additional RF spoiling increments were simulated to match
in vivo acquisitions.
In vivo Acquisitions
Variable
FA data were acquired at 7T with acquisition settings matching the simulations,
a FOV of 192×192×160mm
3 and 1mm isotropic resolution, three RF
spoiling increments (50°, 117°, 144°) and two
nominal B
1+eff (100% and 160%, obtained by
scaling the nominal transmitter voltage). T
1 maps were estimated for
these six conditions using each estimation scheme described for the simulations,
leading to a total of 18 T
1 maps. GM and WM T
1 estimates,
from voxels in which the target B
1eff was obtained, were compared
with simulations.
Results
R2*
and T1 estimates depended on fMW and residency time, with
the sensitivity depending on the estimation scheme (Figure 2). R2*
estimates obtained with the higher FA of 26° showed the greatest variation (6.07%
and 23.70% respectively). While the T1 estimates all varied with fMW
(~30%), the dependence on residency time varied across estimation schemes, with
the ESTATICS approach showing largest sensitivity (4.17%).
T1
decreased with increasing fMW in all cases, regardless of residency
time. B1+eff altered the dependence on
residency time (Figure 3). While
ESTATICS showed greatest variation with B1eff when the residency
time was short (100ms, 9.46% v’s 0.87%), the per-contrast estimates showed
greater variation at longer residency time (500ms, 9.64% v’s 1.36%). The single echo case showed intermediate
sensitivity.
Per-contrast
T1 estimates decreased with residency time for B1eff >100%
but increased for B1eff < 100% (Figure 4). With ESTATICS, T1 estimates
increased with residency time for all B1eff. The single-echo case again
showed intermediate behaviour.
Experimentally
measured T1 times showed the same trends across estimation schemes
as simulations carried out with fMW = 6% and 16% for WM and GM respectively,
a residency time of 300ms, the three RF spoiling increments and two B1+eff
(Figure 5). Discussion
More
rapid relaxation was observed with increasing fMW as expected. However,
the exact dependence varied across estimation schemes. Highest T1
variation was observed with the per-contrast approach, likely driven by the larger
sensitivity to the variable weighting of the sub-compartments across FA.
Residency
time showed a lesser, though still appreciable (Figure 4), effect on T1,
which interacted with B1eff with the per-contrast and ESTATICS
schemes showing opposite behaviour. The behaviour of the single-echo approach was
intermediate. However, the exact
behaviour will depend on the choice of TE: T1 variations converge on the per-contrast
scheme at short TE, but on ESTATICS-like behaviour at long TE (data not shown).
The
dependence of the calculated T1 on how S1 and S2
are estimated was also observed in vivo. Generally, when B1eff increased T1
estimates derived with ESTATICS increased, whereas with the per-contrast scheme
they decreased. While many parameters had to be assumed (e.g. residency time of
300ms, fMW of 6 and 12% for GM and WM respective), there was broad
agreement between the simulations and experiments, plausibly suggesting an intermediate
exchange regime in vivo. Conclusion
Great
care must be taken when assuming a single compartment with variable FA-based T1
estimates. This simplification not only impact R2*
estimates, but propagates through to T1 leading to variability that
is observable in vivo. Furthermore, this variance depends on tissue
features (e.g. fMW, residency time), hardware settings (B1eff)
and sequence choices (notably TE and echo spacing).Acknowledgements
The
Wellcome Centre for Human Neuroimaging is supported by core funding from the
Wellcome [203147/Z/16/Z].References
1. Chan K-S, Marques J. Multi-compartment relaxometry and
diffusion informed myelin water imaging – promises and challenges of new
gradient echo myelin water imaging methods. Neuroimage 2020;221:117159. doi:
10.1016/j.neuroimage.2020.117159.
2. Weiskopf N, Callaghan MF, Josephs
O, Lutti A, Mohammadi S. Estimating the apparent transverse relaxation time
(R2*) from images with different contrasts (ESTATICS) reduces motion artifacts.
Front. Neurosci. [Internet] 2014;8:1–10. doi: 10.3389/fnins.2014.00278.
3. Malik SJ, Teixeira RPAG, Hajnal J
V. Extended phase graph formalism for systems with magnetization transfer and
exchange. Magn. Reson. Med. [Internet] 2018;80:767–779. doi: 10.1002/mrm.27040.
4. Mohammadi S, D’Alonzo C, Ruthotto
L, Polzehl J, Ellerbrock I, Callaghan MF, Weiskopf N, Tabelow K. Simultaneous adaptive
smoothing of relaxometry and quantitative magnetization transfer mapping. 2017.
doi: 10.20347/WIAS.PREPRINT.2432.
5. Labadie C, Lee J-H, Rooney WD,
Jarchow S, Aubert-Frécon M, Springer CS, Möller HE. Myelin water mapping by
spatially regularized longitudinal relaxographic imaging at high magnetic
fields. Magn. Reson. Med. [Internet] 2014;71:375–87. doi: 10.1002/mrm.24670.