Maame Owusu-Ansah1, Candace C. Fleischer1, and Daniel E. Harper2
1Department of Radiology and Imaging Sciences, Emory University School of Medicine, Atlanta, GA, United States, 2Department of Anesthesiology, Emory University School of Medicine, Atlanta, GA, United States
Synopsis
A
proof-of-concept generalized linear model to characterize the relationships
between brain metabolite concentrations and pain perception is presented,
explicitly accounting for differences in regional grey matter (GM) density. Brain
metabolite concentrations measured with magnetic resonance spectroscopy (MRS),
and pain perception characterized using quantitative sensory testing (QST), were
significantly associated in a cohort of patients with chronic pain. These
results contribute to the expanding literature that supports the utility of neuroimaging
to understand the underlying mechanisms of centralized chronic pain.
Introduction
Previous
research to identify neuroimaging signatures of chronic pain has led to varied
results. While morphological brain changes may contribute to the persistence
and evolution of chronic pain, numerous reports have observed decreases while
others observe increases in GM density as a function of chronic pain [1,2]. Similarly,
the use of MRS to characterize neurochemical abnormalities, such as the dysregulation
of neurotransmitters including glutamate, is an increasingly common approach.
However, both increases and decreases in brain metabolites have been reported,
resulting in a limited understanding of the role of brain metabolites in pain
perception. Taken together, there is a clear need for improved non-invasive MR
biomarkers for prognostication and treatment monitoring to further explore the
underlying mechanisms that drive chronic centralized pain. Here, we present a
proof-of-concept approach using a generalized linear model to characterize
changes in brain metabolites in a cohort of temporomandibular disorder (TMD) patients
with chronic pain.Methods
A
schematic of the study design is shown in Figure 1. After providing written
informed consent, MR data and quantitative sensory testing (QST) metrics were
acquired in TMD patients (n=16) and healthy controls (HCs; n=12). MR data was
acquired on a 3T MR scanner (GE Signa) using a 32-channel receive array head
coil. T1-weighted SPGR (TR/TE = 650/3.7
ms, flip angle = 8°, voxel resolution = 0.5x0.5x0.4 mm3, NEX = 1)
and single-voxel PRESS MRS (TR/TE = 1500/35 ms, flip angle = 90°,
complex data points = 4096, 2 cm isotropic voxel in the anterior cingulate
cortex, ACC) were acquired. MR spectra were analyzed with LCModel (v6.3-1L) [3]. Tissue metabolite concentrations
were corrected for cerebrospinal fluid (CSF) after segmentation with FSL-FAST [4]. FSL-VBM
(http://www.fmri-b.ox.ac.uk/fsl/) [5-7] was used to quantify GM morphometry in
the ACC region using the following steps: 1) Brain tissue was extracted using
BET [8]. 2) A study-specific template was formed using a randomly chosen subset
of subjects (11 HCs and 11 TMD patients) and nonlinear registration using FNIRT,
image averaging, tissue segmentation, and affine registration to the MNI152
template using FLIRT [9]. 3) Individual GM images were
nonlinearly registered to the template and modulated with a Jacobian of the
warp field. 4) These images were then smoothed with a Gaussian kernel with
sigma of 7 mm. 6) Finally, an ACC region of interest (ROI) mask was generated
and used to calculate regional GM density (mm3) for each subject
using the individual images that were registered to the study-specific
template. Quantitative sensory testing (QST) metrics were used to quantify pain
perception and included: current face pain as a marker for chronic pain
symptoms; symptom severity using the widespread pain index and fibromyalgia-ness scores; fibromyalgia
determined with the American College of Rheumatology’s 2011 Fibromyalgia
Criteria; pressure algometry in the trapezius and knee muscles; and cuff
algometry using a blood pressure cuff inflated to 160 mm Hg. Statistical
analysis was performed in SPSS v26 (IBM) using generalized linear models.
Significance was determined at p<0.05.Results and Discussion
Descriptive
statistics are shown in Table 1. Significant differences in metrics of
centralized pain including current face pain, fibromyalgia score, and symptom
severity were observed in TMD patients compared to HCs. Age and sex were not
significantly different between groups. Initial analysis included linear regressions
of glutamate+glutamine (Glx) with QST metrics in TMD and HCs (Tables 2,3). Significant
associations of Glx with QST metrics were observed when accounting for GM
density in TMD patients (GLM regressions, Table 2). Inclusion of GM density in
the ACC ROI as an additional linear covariate in our model resulted in improved
model fits for TMD patients. In comparison, Glx was not significantly associated
with most QST metrics in healthy controls (Table 3). Brain metabolite
concentrations are known to vary as a function of GM volume within and between
subjects [10]. As a result, GM volume or density must be considered as a
potential covariate when characterizing changes in metabolites, particularly as
GM volumes have been observed to significantly vary with symptom severity. Significant
model associations between Glx and pain sensitivity in the knee and trapezius
muscles using algometry were observed, suggesting that dysregulation of brain
metabolites may be associated with heightened pain perception. While some
studies that account for differences in GM use average GM concentrations as a
correction factor, or MRS voxel segmentation to quantify GM fractions, the use
of morphometry is an alternative correction approach presented here. The
benefit of incorporating GM density from morphometry, rather than GM fractions
from voxel segmentation, is that morphometry accounts for differences in brain
size and local concentration variations between subjects and does not assume GM
metabolite concentrations a priori. Importantly, previous discrepancies in
reported metabolite concentrations may be attributed to variations in GM morphometry. Explicitly including GM density as a covariate in the model can account for tissue
differences at the subject level. Conclusion
Improved
predictions of the association between brain metabolites and pain perception
was facilitated with a generalized linear model that accounts for differences
in GM density. Morphometry may be a promising alternative to voxel segmentation
to perform tissue composition corrections when quantifying brain metabolite
changes in chronic pain.Acknowledgements
This work
was supported in part by NIH K99DE026810. References
- Smallwood, RF et al. Structural brain anomalies and chronic pain: a
quantitative meta-analysis of gray matter volume. J Pain. 2013;14(7):663-75.
- Cauda, F et al. Grey matter alterations in chronic pain: A
network-oriented meta-analytic approach. Neuroimage Clin. 2014;4:676-686.
- Provencher,
SW. Estimation of metabolite concentrations from localized in vivo proton NMR
spectra. Magn Reson Med. 1993;30:672-379
- Zhang Y, Brady, M, Smith S. Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm. IEEE Trans Med Imaging. 2001;20:45–57.
- Douaud, G
et al. Anatomically related grey and white matter abnormalities in
adolescent-onset schizophrenia. Brain. 2007;130:2375-2386.
- Good, CD
et al. A voxel-based morphometric study of ageing in 465 normal adult human
brains. Neuroimage. 2001;14:21-36.
- Smith,
SM et al. Advances in functional and structural MR image analysis and
implementation as FSL. Neuroimage. 2004;23(Suppl 1):S208-219.
- Smith, SM. Fast robust automated brain extraction. Hum Brain Mapp, 2002;17(3):143-55.
- Jenkinson, M and Smith, S, A global optimisation method for robust
affine registration of brain images. Med Image Anal. 2001;5(2):143–156.
- Tal, A et
al. The role of grey and white matter segmentation in quantitative proton MR
spectroscopic imaging. NMR Biomed. 2013;25(12):1392-1400.