Elsevier

NeuroImage

Volume 182, 15 November 2018, Pages 370-378
NeuroImage

Impact of magnetic susceptibility anisotropy at 3 T and 7 T on T2*-based myelin water fraction imaging

https://doi.org/10.1016/j.neuroimage.2017.09.040Get rights and content

Highlights

  • Estimation of the impact of the susceptibility of myelin on the MGRE-MWF.

  • Myelin's susceptibility may not need to be considered for MGRE-MWF imaging at 3 T.

  • Myelin's susceptibility should be considered for MGRE-MWF imaging at 7 T.

Abstract

Purpose

Myelin Water Fraction (MWF) mapping can be achieved by fitting multi-gradient recalled echo (MGRE) magnitude images with a three-component model or a pseudo-continuous T2 distribution. Recent findings of compartment-specific orientation-dependent magnetic susceptibility shifts have spurred the inclusion of frequency offset (Δf) terms in the fitting models. In this work, we performed simulations to assess the impact of Δf's on the MWF, derived from three different fitting models, at two field strengths.

Theory and methods

White matter MGRE signals were simulated using the Hollow Cylinder Fiber Model at 3 and 7 T, for a range of fiber orientations (θ), and analyzed using: 1) a multi-component T2 signal magnitude model (MCMT2); 2) a three-component T2 signal magnitude model (3CMT2); and, 3) a three-component complex T2 signal model (3CCT2).

Results

At 3 T, MCMT2 & 3CMT2 yielded accurate MWFs: (11.9±1.1)% and (11.7±1.0)% (mean± standard deviation across 1000 simulations, true MWF = 12%), respectively. 3CCT2 MWFs were less accurate and had the largest variability: (9.2±5.0)%. At 7 T, MCMT2 and 3CMT2 MWFs became less accurate as θ increased. This was remedied by 3CCT2, at the expense of accuracy for small θ.

Conclusion

This work suggests that if no information regarding Δf is sought, MCMT2 and 3CMT2 are preferable at 3 T. At 7 T, Δf cannot be overlooked.

Introduction

Quantitative myelin-specific imaging using MRI has been pursued for many years with a variety of techniques. One approach for quantifying myelin is based on analysis of T2 relaxation. The multicomponent nature of transverse relaxation in central nervous system tissue has been established in a number of studies (Vasilescu et al., 1978, Peled et al., 1999, Does and Snyder, 1996, Stewart et al., 1993, Menon and Allen, 1991) using detailed nuclear magnetic resonance relaxation measurements. At least three distinct components have been detected, and attributed to: water trapped between the lipid bilayers of myelin (i.e. myelin water (MW)), axonal water (AW), and extracellular water (EW). MacKay et al. (1994) introduced an in vivo technique based on a single-slice multi-echo spin-echo (MESE) CPMG sequence to acquire T2 relaxation data, with subsequent multicomponent analysis. Two sizable T2 components (T2 of 10–40 ms and T2 of 70–100 ms) were identified and attributed to MW and to the combination of AW and EW (collectively referred to as intra/extracellular water (IEW)) (Does and Gore, 2002), respectively. The data were analyzed using a regularized non-negative least squares (NNLS) algorithm (Whittall and MacKay, 1969) to fit a T2 distribution to the magnitude MESE data. Moreover, the myelin water fraction (MWF) was defined as the ratio of the area under the MW peak to the total water signal. Strong correlations have since been reported between the MWF and histological measures of myelin (Webb et al., 2003, Laule et al., 2006, Laule et al., 2008), validating the use of the MWF as a biomarker for myelin content in central nervous system tissue. MWF imaging, based on the aforementioned single slice MESE sequence, is not clinically feasible due to its limited spatial coverage and long scan time. However, this technique has recently been extended to 3D using a combined gradient and spin echo sequence (GRASE) (Prasloski et al., 2012a). Multicomponent analysis of GRASE T2 data is based on regularized NNLS with concurrent correction for stimulated echo contamination of the signal decay curve (Prasloski et al., 2012b).

Alternate multi-slice and 3D approaches to MWF imaging that have been proposed include: T2-preparation-based imaging (Oh et al., 2006, Nguyen et al., 2012), multicomponent driven-equilibrium single-pulse observation of T1 and T2 (mcDESPOT) (Deoni et al., 2008), and MWF mapping based on multi-gradient recalled echo (MGRE) imaging with multicomponent analysis of the magnitude T2 decay curve (Du et al., 2007, Hwang et al., 2010, Lenz et al., 2012). MGRE imaging has the advantage of offering fast whole brain coverage, with low specific absorption rate (SAR), and short first echo time (TE1) and echo spacing (ES). Two methods have been used to analyze MGRE data: regularized NNLS fitting of a pseudo-continuous multi-component T2 distribution (Lenz et al., 2012) and a three-component model (Andrews et al., 2005, Lancaster et al., 2003). In the latter, the three components are assumed to be: MW, EW, and AW, and there are 6 free parameters (3 relaxation times and 3 component amplitudes).

Recently (van Gelderen et al., 2012), a three-component model with additional free parameters, corresponding to frequency offsets (Δf) for the water components, was found to better fit magnitude MGRE data at 7 T. The Δf’s appeared to depend on both field strength and the orientation (θ) of WM fiber bundles relative to the main magnetic field (B0). More specifically, the Δf for MW was found to be largest in fibers that were oriented perpendicular to the B0 field (identified from diffusion tensor imaging (DTI) (Le Bihan et al., 2001) fiber orientation maps). This anisotropic behavior of Δf has been attributed to anisotropy in the magnetic susceptibility of myelin, based on the highly ordered structure of the membranes making up the myelin sheath (Sati et al., 2013, He and Yablonskiy, 2009). When fitting complex MGRE signals obtained in human brain at 7 T to a three-component model that includes Δf, Sati et al. (2013) found that when fibers were perpendicular to B0, MW, AW and EW experienced strong positive, negative and near-zero Δf, respectively. Nam et al. (2015) used a similar model at 3 T, and found that perpendicular fibers had positive and negative frequency offsets for MW and AW compartments, whereas parallel fibers had slight positive MW offsets and AW had slight negative offsets. The three-component complex model proposed by Nam et al. (2015) also led to more stable MWF estimates, compared to three-component models that either do not include Δf terms (Hwang et al., 2010), or do include such terms, but involve fitting only magnitude data (van Gelderen et al., 2012). van Gelderen et al. (2012) observed that three-component fits of signals obtained from the splenium of the corpus callosum at 7 T exhibited highly significant residuals (refer to Appendix A in (van Gelderen et al., 2012)). If the presence of large resonant frequency offsets results in MGRE signals that possess significant deviations from exponential behavior, complex fitting of these signals may be necessary for accurate MWF imaging. The inclusion of Δf terms in multicomponent fitting of MGRE signals is also of significant interest because it allows one to obtain structural information related to the orientation of WM fibers that would otherwise not be available when using NNLS or 3-component fitting approaches. Furthermore, the Δf terms would also enable one to distinguish between the AW and EW components in three-component fitting (Sati et al., 2013). These features render MGRE-based MWF imaging more attractive, compared to MESE-based MWF imaging.

In this work, we assessed the impact of orientation-dependent Δf on the MWF derived from multicomponent analysis of simulated T2 data at two field strengths: 3 T and 7 T. We used the Hollow Cylinder Fiber Model (HCFM), recently proposed by Wharton and Bowtell (2012), to simulate complex multicomponent MGRE magnitude and phase signals in WM. Next, we analyzed these signals using three different methods: regularized NNLS multi-component fitting of a T2 distribution to the signal magnitude (MCMT2), three-component T2 fitting to the signal magnitude (3CMT2), and three-component T2 fitting with Δf terms to the complex T2 signals (3CCT2). To validate our findings, we acquired MGRE data in 9 healthy volunteers at 3 T. 6 volunteers were scanned once with the MGRE sequence to compare the three fitting methods in-vivo, and 3 were scanned twice (during the same scan session) with the MGRE sequence to evaluate the stability of the three fitting methods.

Section snippets

Simulations

Multicomponent T2 decay was simulated for white matter at 3 T and 7 T. T2 components were generated for MW, EW, and AW, with relaxation times for data simulated at 3 T equal to: 10 ms, 48 ms, and 64 ms (Nam et al., 2015), respectively. At 7 T, the relaxation times were correspondingly set to 6 ms, 30 ms, and 40 ms (Sati et al., 2013). Since there is a lack of consensus in the literature as to which of the longer T2 component belongs to EW and which belongs to AW (Hwang et al., 2010, Sati

Simulations

At 3 T, both MCMT2 and 3CMT2 led to accurate MWF values, independent of the assumed orientation of the myelin fiber. When considering all simulated angles (200 repetitions x 5 angles = 1000 simulations), with SNR = 140 and true MWF = 12%, the mean MWF (±standard deviation (std)) was (11.9 ± 1.1)% and (11.7 ± 1.0)% for MCMT2 and 3CMT2, respectively. 3CCT2 led to a slight increase in MWF variability: (9.2 ± 5.0)%, across all 1000 simulations. Fig. 1 shows the median MWF ± the interquartile

Discussion and conclusions

In this work we evaluated the impact of fiber orientation-dependent frequency offsets on the MWF derived from multicomponent T2 analysis of MGRE signals. First, realistic multicomponent MGRE signals originating from WM were simulated using the HCFM model. Second, the simulated signals were analyzed using three different methods: MCMT2 and 3CMT2, which do not include Δf terms, and 3CCT2, which does include Δf terms. Third, MWFs were computed from the fits obtained using the three methods and

Acknowledgments

This study was supported by the Fonds de Recherche du Québec - Nature et Technologies (FRQNT), Graduate Scholarship to Eva Alonso Ortiz, the CREATE Medical Physics Research Training Network grant of the Natural Sciences and Engineering Research Council (NSERC, Grant number 432290), the Canadian Institutes of Health Research (CIHR, Funding Reference Number 43871), and the Natural Sciences and Engineering Research Council (NSERC Discovery, Grant number 170426). We would like to thank Raphael

References (47)

  • E. Alonso-Ortiz et al.

    Field inhomogeneity correction for gradient echo myelin water fraction imaging

    Magn. Reson Med.

    (2016)
  • E. Alonso-Ortiz et al.

    Multi-gradient-echo myelin water fraction imaging: comparison to the multi-echo-spin-echo technique

    Magn. Reson Med.

    (2017)
  • T. Andrews et al.

    Testing the three-pool white matter model adapted for use with T2 relaxometry

    Magn. Reson Med.

    (2005)
  • P. Coupe et al.

    An optimized blockwise nonlocal means denoising filter for 3-D magnetic resonance images

    IEEE Trans. Med. Imaging

    (2008)
  • S. Deoni et al.

    Gleaning multicomponent T1 and T2 information from steady-state imaging data

    Magn. Reson Med.

    (2008)
  • M. Does et al.

    Compartmental study of T1 and T2 in rat brain and trigeminal nerve in vivo

    Magn. Reson Med.

    (2002)
  • M.D. Does et al.

    Multiexponential T2 relaxation in degenerating peripheral nerve

    Magn. Reson Med.

    (1996)
  • Y. Du et al.

    Fast multislice mapping of the myelin water fraction using multicompartment analysis of T2* decay at 3T: a preliminary postmortem study

    Magn. Reson Med.

    (2007)
  • S.J. Graham et al.

    Criteria for analysis of multicomponent tissue T2 relaxation data

    Magn. Reson Med.

    (1996)
  • H. Gudbjartsson et al.

    The Rician distribution of noisy MRI data

    Magn. Reson Med.

    (1995)
  • X. He et al.

    Biophysical mechanisms of phase contrast in gradient echo MRI

    Proc. Natl. Acad. Sci.

    (2009)
  • A. Istratov et al.

    Exponential analysis in physical phenomena

    Rev. Sci. Instrum.

    (1999)
  • D. Kim et al.

    Probing signal phase in direct visualization of short transverse relaxation time component (ViSTa)

    Magn. Reson Med.

    (2015)
  • Cited by (17)

    • Signal compartments in ultra-high field multi-echo gradient echo MRI reflect underlying tissue microstructure in the brain

      2018, NeuroImage
      Citation Excerpt :

      Previous studies which use the three-compartment water fraction model to explain temporal trends in mGRE-MRI data have been confined to white matter regions (Li et al., 2015; Nam et al., 2015; Sati et al., 2013; Wu et al., 2017). Frequency shifts in ‘extracellular’ fluid and CSF are commonly assumed to be zero (Alonso-Ortiz et al., 2017; Nam et al., 2015; van Gelderen et al., 2012; Wu et al., 2017). A recent study performed by Straub et al. (2017) found the CSF to be a stable and robust region for susceptibility referencing in healthy participants; the authors report a volume susceptibility of 0.010 ppm within this region, corresponding to a frequency shift of 3.0 Hz at 7T after converting in accordance with Eq. (1).

    View all citing articles on Scopus
    1

    Present Address: Department of Medical Physics, The Ottawa Hospital Cancer Centre, 501 Smyth Rd., Ottawa, Ontario, K1H 8L6, Canada.

    View full text