Chemical shift-based MRI to measure fat fractions in dystrophic skeletal muscle

Abstract

Purpose: The relationship between fat fractions (FFs) determined based on multiple TE, unipolar gradient echo images and (1) H magnetic resonance spectroscopy (MRS) was evaluated using different models for fat-water decomposition, signal-to-noise ratios, and excitation flip angles.

Methods: A combination of single-voxel proton spectroscopy ((1) H-MRS) and gradient echo imaging was used to determine muscle FFs in both normal and dystrophic muscles. In order to cover a large range of FFs, the soleus and vastus lateralis muscles of 22 unaffected control subjects, 16 subjects with collagen VI deficiency (COL6), and 71 subjects with Duchenne muscular dystrophy (DMD) were studied. (1) H-MRS-based FF were corrected for the increased muscle (1) H2 O T1 and T2 values observed in dystrophic muscles.

Results: Excellent agreement was found between coregistered FFs derived from gradient echo images fit to a multipeak model with noise bias correction and the relaxation-corrected (1) H-MRS FFs (y = 0.93x + 0.003; R(2) = 0.96) across the full range of FFs. Relaxation-corrected (1) H-MRS FFs and imaging-based FFs were significantly elevated (P < 0.01) in the muscles of COL6 and DMD subjects.

Conclusion: FFs, T2 , and T1 were all sensitive to muscle involvement in dystrophic muscle. MRI offered an additional advantage over single-voxel spectroscopy in that the tissue heterogeneity in FFs could be readily determined.

Keywords: Duchenne muscular dystrophy; collagen VI; congenital muscular dystrophy; fat water imaging; magnetic resonance spectroscopy; muscle composition; skeletal muscle.

Copyright © 2013 Wiley Periodicals, Inc.

Figures

Figure 1

Example of co-registration of single voxel MRS data from the soleus muscle with multislice water maps generated by MRI from calf of a DMD subject in (a) sagittal orientation and (b) axial orientation. A typical muscle spectrum is shown in (c) with numbered peaks corresponding to Table 1.

Figure 2

Phantom testing of water/fat decomposition algorithm using a single peak (SP) and a multi-peak (MP) spectral models on three individual phantoms. Scan plan location (a) and fat fraction map (b) with 1H-MRS voxel location shown in red. (c) Relationship between MRI derived fat fractions and localized single voxel 1H spectra. The dash line represents the line of identity. Standard deviations are represented by error bars inside data points.

Figure 3

Relationship between T1 and T2 values from the SOL of control and DMD subjects. We found a linear relationship (R2=0.72) between T1 and T2, represented by the solid line. The dashed lines represent the 95% confidence interval of the fit.

Figure 4

Relationship between muscle fat fraction measured by MRI and MRS in all groups and muscles using (a) a single peak model (y=0.74×−0.02, R2=0.93), or (b) a multi-peak (8-peak) model (y=0.96×+0.01, R2=0.98) for fat spectral decomposition. MRS determined fat fractions were determined based using a model independent method (peak integration in a and b) and using a 8 peak fat-water model in AMARES (y=0.93×+0.02, R2=0.98) (c). The dashed line represents the identity line. The corresponding Bland-Altman plots are shown to the right of each relationship.

Figure 5

Effect of the noise bias correction at high (20° flip angle) and low (5° flip angle) signal to noise ratio in the soleus. (a) MRI based FF with no noise bias correction as a function of MRS FF. The relationship showed a deviation from linearity at low FFs. (b) Noise bias corrected MRI FF based on the 20° flip angle acquisition and (c) based on the 5° flip angle acquisition, as a function of MRS FF. The dashed line represents the identity line.

Figure 6

Muscle fat fraction measured in CON, COL6, DMD groups in soleus by (a) MRS and (b) MRI, and in the vastus lateralis by (c) MRS and (d) MRI.

Figure 7

Examples of histograms from the medial gastrocnemius of two different COL6 subjects with similar fat fraction as measured by MRI. Subject (a) showed fatty infiltration inward from the external fascia, which was associated with a bimodal fat fraction distribution (c), whereas subject (b) showed a unimodal distribution characteristic of a more diffuse pattern of fat infiltration.