Synopsis
Multiexponential T2 and quantitative magnetization transfer experiments provide quantitative measures of myelin water fraction (MWF) and bound pool fraction (BPF), respectively. These measures are known to correlate with myelin content in white matter; however discrepancies between the two have been shown. We display that by correcting for all proton pools contributing to MWF and BPF in white matter, we are able to show similar absolute measures of myelin content from MWF and BPF that are nearly equal to each other and close to myelin content measured by quantitative histology.Purpose
Myelin water fraction (MWF) derived from multiexponential T
2
(MET
2) and bound pool fraction (BPF) resulting from quantitative
magnetization transfer (qMT) experiments are known to correlate with myelin
content in white matter (1,2).
However, comparative studies have shown discrepancies between these two
measures (3–5), which
raises the question of how exactly these quantitative measures should be
interpreted. Here we test the relationships between MET
2, qMT, and
myelin volume fraction (as measured by electron microscopy), in the context of
a four-pool model of protons in white matter.
Methods
Mouse brains from control (n=6)
and two models of tuberous sclerosis complex (Rictor CKO, n=3 and TSC, n=3)
were perfusion-fixed and loaded with 1mM of Gadolinium (Magnevist) for high
resolution 3D Extended Phase Graph (EPG)-compliant MET2 and
inversion-recovery prepared 3D fast spin echo qMT studies at 15.2 T with
150µm isotropic resolution for both. Myelin water fraction (MWF) was obtained
from voxel-wise T2-spectra fitting of MET2 data and bound
pool fraction (BPF) was extracted voxel-wise from qMT data as previously
described (6). Histology was performed using transmission electron microscopy on 4
regions of white matter from each brain. Histologic images were carefully
segmented to obtain measures of myelin volume fraction (fHIST), as shown in Fig 1.
We use a four-pool model (Fig 2)
to describe the fractions of water and macromolecular protons in non-myelin (a and b, respectively) and myelin (c
and d, respectively) tissues. We
assume approximately equal proton densities in each pool (7) and a + b + c + d = 1, therefore
the myelin volume fraction = c + d. Assuming no exchange of magnetization
between the myelin and non-myelin proton pools, MWF = d/(b+d), and from the standard binary spin pool qMT model, assuming identical biexponential recovery rates of longitudinal magnetization, BPF = a+c.
We then define fMET2 and fMT as measures of myelin
volume fraction (i.e., = c + d) which
are expressible in terms of MWF and BPF, respectively.
Using literature values of water
mass fractions in myelin (ρm = d/(c+d) = 0.44) and non-myelin (ρn= b/(a+b)
= 0.82) components of white matter (8), MWF can be written in terms of the myelin volume fraction (fMET2 = c+d) as
shown in Eq. II, which can then be solved
to express fMET2 as a
function of MWF as shown in Eq. III.
Similarly for qMT analysis, we first
use the linear correlation between BPF and fHIST
(Fig 3), to determine the contribution of the non-myelin macromolecular proton
fraction (a) to BPF. Then, using water
mass fraction in myelin, ρm, again, the myelin volume fraction (fMT = c+d) can be expressed in terms of BPF as shown in Eq. V.
Results
Fig 1 displays qualitative decreases in
fHIST in representative segmented myelin images from
control, Rictor CKO, and TSC brains. Fig 3 (top) displays MWF (black x) and BPF
(blue dot) versus
fHIST
with BPF offset (
a) and the
line of unity (gray, dashed). While all measures display good correlation, with gradual decreases in myelin
content, as expected in models, the absolute values of MWF, BPF and
fHIST do not agree. Fig 3 (bottom) shows
fMET2 (black x) and
fMT
(blue dot) versus
fHIST,
which shows strong linear correlation between
fMET2 and
fMT
(R = 0.89) and each to
fHIST (R =
0.8 and 0.83 for
fMET2
and
fMT, respectively). Additionally,
fMET2,
fMT, and
fHIST exhibit good agreement in absolute measures of
myelin content, with
fMET2
and
fMT displaying nearly
identical results. However, it is apparent that both
fMET2 and
fMT
underestimate
fHIST, which
may reflect a systematic overestimation of
fHIST
from histological image analysis, or, for
fMET2
at least, this could be explained by intercompartmental water exchange (5,9).
Conclusion
By correcting for all proton pools contributing to MWF and
BPF, we are able to show similar absolute measures of myelin content (
fMET2 and
fMT) that are nearly equal to
each other and close to myelin content measured by quantitative histology (
fHIST). This displays that
BPF can be as specific to myelin content as MWF, which is important because BPF
is generally a higher precision measurement. However, if accurate measures
of myelin volume fraction rather than simply correlative measures are needed,
for example for measuring g-ratio (10,11) conversion
of BPF to
fMT relies on
accounting for the contribution from non-myelin macromolecular protons, which
may not be the same across all tissues or subjects.
Acknowledgements
NIH EB001744 and NSF GRFP DGE-0909667References
1.
Laule C, Leung E, Li D, Traboulsee AL. Myelin water imaging in multiple
sclerosis: quantitative correlations with histopathology. Multiple Sclerosis
2006;12:747–753.
2.
Schmierer K, Tozer D, Scaravilli F, Altmann D, Barker G, Tofts P, Miller D.
Quantitative magnetization transfer imaging in postmortem multiple sclerosis
brain. J Magn Reson Imaging 2007;26:41–51. doi: 10.1002/jmri.20984.
3.
Sled JG, Levesque I, Santos AC, Francis SJ, Narayanan S, Brass SD, Arnold DL,
Pike GB. Regional variations in normal brain shown by quantitative
magnetization transfer imaging. Magn. Reson. Med. 2004;51:299–303. doi:
10.1002/mrm.10701.
4.
Tozer DJ, Davies GR, Altmann DR, Miller DH, Tofts PS. Correlation of apparent
myelin measures obtained in multiple sclerosis patients and controls from
magnetization transfer and multicompartmental T2 analysis. Magn. Reson. Med.
2005;53:1415–1422. doi: 10.1002/mrm.20479.
5.
Dula A, Gochberg D, Valentine H, Valentine W, Does M. Multiexponential T2,
magnetization transfer, and quantitative histology in white matter tracts of
rat spinal cord. Magnetic Reson. Medicine 2010;63. doi: 10.1002/mrm.22267.
6.
Gochberg D, Gore J. Quantitative imaging of magnetization transfer using an
inversion recovery sequence. Magn Reson Med 2003;49:501–505. doi: 10.1002/mrm.10386.
7.
Horch AR, Gore JC, Does MD. Origins of the ultrashort T21H NMR signals in
myelinated nerve: A direct measure of myelin content? Magnetic Resonance in
Medicine 2011;66:24–31. doi: 10.1002/mrm.22980.
8.
Knaap M, Valk J. Magnetic Resonance of Myelination and Myelin Disorders. 3rd
ed. Springer-Verlag Berlin Heidelberg; 2005 pp. 6–7. doi:
10.1007/3-540-27660-2_1.
9.
Levesque IR, Pike GB. Characterizing healthy and diseased white
matter using quantitative magnetization transfer and multicomponent T2
relaxometry: A unified view via a four-pool model. Magnetic Resonance in Medicine
2009;62:1487–1496. doi: 10.1002/mrm.22131.
10.
Stikov N, Campbell J, Stroh T, et al. In vivo histology of the myelin g-ratio
with magnetic resonance imaging. NeuroImage 2015;118:397405. doi:
10.1016/j.neuroimage.2015.05.023.
11.
West KL, Kelm ND, Carson RP, Does MD. A revised model for estimating g-ratio
from MRI. Neuroimage 2015. doi: 10.1016/j.neuroimage.2015.08.017.