PHN patient recruitment

Right-handed patients diagnosed with PHN by two physicians (XH and KM) were recruited at Xin Hua Hospital in Shanghai, China. Patients who reported a history of shingles and associated pain and had had PHN symptoms >3 months were included in the study. Patients of age <50 or >75, with psychiatric diseases, with claustrophobia, current users of opioids, or with metal or electronic implants were excluded. Patients were using only nonsteroidal anti-inflammatory (eg, celecoxib) and/or peripheral neurotrophic drugs when they were recruited.

Pain assessment in PHN patients

To minimize the potential acute analgesic effect of the nonsteroidal anti-inflammatory drug that the PHN patients had taken, patients were asked not to take any medicine for at least 24 hours prior to the experiment, a washout period that allowed about 90% elimination of plasma celecoxib²⁶ but avoided prolonged halt of medicine, as adopted in some other studies with chronic pain patients.^27–29 PHN patients were asked to report their perceived spontaneous pain on a commonly used pain numeric rating scale³⁰ of 0–10 (0 for no pain and 10 for worst imaginable pain) right before the functional magnetic resonance imaging (fMRI) scan (Table 1). Pain score over the last 7 days prior to the fMRI experiment was also assessed in the patients.

Imaging data acquisition

Imaging was performed on a 3 T Trio scanner (Siemens, Munich, Germany) using an eight-channel birdcage head coil. Each patient lay supine with their head snugly fixed by foam pads. The patient was asked to keep still as long as possible and to keep his/her eyes closed but remain awake. Resting-state fMRIs were obtained using an echo-planar imaging sequence with protocols of 30 axial slices, thickness/ gap 4.0/0.8 mm, in-plane resolution 64×64, TR 2000 ms, TE 40 ms, flip angle 90°, FOV 220×220 mm, 240 volumes acquired in 8 minutes.

Imaging data preprocessing

Preprocessing of resting-state fMRI data was conducted using Data Processing Assistant for Resting-State fMRI (DPARSF; http://rfmri.org/dparsf) software (version 2.1) based on statistical parametric mapping (SPM8; http://www.fil.ion.ucl.ac.uk/spm) on the MatLab platform (MathWorks, Natick, MA, USA).³¹ In brief, the first ten volumes of each functional time series were discarded, as patients were adjusting themselves during that period to the fMRI environment. The remaining 230 images were slice-time-corrected with the 30th slice as the reference and spatially realigned for head motion. Patient head motion was assessed by evaluating three translations and three rotations for each scan. Translational thresholds were set to ±2 mm, while rotational thresholds were limited to ±1°. After head motion correction, functional images were spatially normalized to Montreal Neurological Institute (MNI) space using echo-planar imaging sequence templates (resampled voxel size 3×3×3).

Regional homogeneity

For ReHo measurements,²³ normalized images were further analyzed using DPARSF.³¹ First, we employed it to remove linear trends and reduce low-frequency drift and high-frequency respiratory and cardiac noise using a temporal band-pass (0.01–0.08 Hz) filter. Subsequently, we used it to map regional spontaneous activity across the whole brain, with Kendall’s coefficient of concordance used for assessing in a voxel-wise fashion the similarity of the time series at a given voxel to the time series of its 26 nearest neighbors:

where R_i is the sum rank of time point i, $\overline{\mathrm{R}}=\frac{[(\mathrm{n}+1)\mathrm{K}]}{2}$ is the mean of R_i values, K is the number of time series corresponding to the spatial voxels contained within a measured cluster (here, K=27, one given voxel plus the number of its neighbors), n is the number of ranks (here, n=230 time points), and i is from 1 to n. Kendall’s coefficient values (W) ranging from 0 to 1 were calculated voxel by voxel across the whole brain, producing an individual ReHo map for each participant. For the purpose of standardization, each individual ReHo map was divided by its own mean ReHo of that entire brain. Then, standardized ReHo maps for all participants were smoothed using a Gaussian filter of 8 mm full width at half maximum to reduce noise and residual differences in normalization.

A total of 26 frontal cortical areas were selected from anatomical automatic labeling,³² as listed in Table 2, to compose a mask used for the ReHo analysis (Figure 1). Within the frontal lobe, we calculated Pearson’s correlation between ReHo of each voxel and subjective spontaneous pain reported by patients right before the scan. The threshold for significance and correction for multiple comparisons were calculated using AlphaSim program (https://afni.nimh.nih.gov/pub/dist/doc/manual/AlphaSim.pdf). A corrected threshold of P<0.05 was utilized with a combined cutoff threshold of P<0.01 and a minimum cluster size of 1,161 mm³ (43 voxels), determined by 1,000 iterations of Monte Carlo simulation. The resulting r map was overlaid on rendering views with BrainNet Viewer (http://www.nitrc.org/projects/bnv), and the anatomy of surviving brain regions was reported using xjView software (http://www.alivelearn.net/xjview).

Open in a separate window

Figure 1

Frontal brain areas whose ReHo was correlated with PHN pain.

Notes: Upper left: mask of frontal lobe. The mask (yellow) consists of 26 anatomical brain areas from anatomical automatic labeling. Names of the 26 anatomical brain areas are listed in Table 2. Upper right: brain areas whose ReHo was correlated with spontaneous pain perceived by PHN patients (P<0.05, corrected). Color bar indicates r-score and cold color indicates negative correlation. Lower: correlations between the ReHo signal from regions of interest and pain scores.

Abbreviations: L, left; OFC, orbitofrontal cortex; PFC, prefrontal cortex; PHN, postherpetic neuralgia; PNRS, pain numeric rating scale; R, right; smReHo, smoothened regional homogeneity.

Functional connectivity

For measurements of functional connectivity,²⁵ fMRI data from preprocessing were further analyzed using DPARSF.³¹ The preprocessed data were smoothed using a Gaussian filter of 8 mm full width at half maximum to reduce noise and residual differences in normalization. Then, the linear trend was removed and a temporal band-pass (0.01–0.08 Hz) filter applied to reduce low-frequency drift and high-frequency respiratory and cardiac noise. Pearson’s correlation was calculated between averaged time courses of regions of interest (ROIs; clusters where ReHo correlated with patients’ spontaneous pain) for each patient in a cluster-wise manner across the whole brain, with global mean time course, white-matter mean time course, cerebrospinal fluid mean time course, and the six head motion parameters as nuisance covariates. Individual correlation coefficients were converted to z-scores using Fisher’s transformation to improve normality. Pearson’s correlation between functional connectivity of each voxel across the whole brain and patients’ subjective spontaneous pain was calculated. A corrected threshold of P<0.05 was utilized with a combined cutoff threshold of P<0.01 and a minimum cluster size of 1,917 mm³ (71 voxels), determined by 1,000 iterations of Monte Carlo simulation. The resulting r-maps were overlaid on rendering views with BrainNet Viewer and on axial slice views with the Rest slice viewer (http://restfmri.net/forum/index.php), and the anatomy of survived brain regions were reported using xjView software.

Signal extraction from ROI

We extracted averaged signals of ReHo or functional connectivity from ROIs for illustrating correlations with PHN pain with the Resting-State fMRI Data Analysis Toolkit version 1.8.³³ Specifically, we first saved surviving clusters from the analysis of ReHo or functional connectivity into binary masks. Then, intersections were examined from these masks and related anatomical regions from the anatomical automatic labeling³² template to obtain particular masks of clusters of interest. Finally, mean values of voxels within each cluster of interest were extracted.

Relationships between prefrontal local synchronization and pain intensity

To explore whether neural activity in the frontal lobe was associated with PHN pain, we examined the relationship between the degree of local synchronization (measured by ReHo) of the PFC and the intensity of pain experienced by PHN patients. ReHo values within the frontal lobe mask consisting of 26 areas³² were analyzed in these PHN patients. Significant negative correlations were observed between pain scores and ReHo values in several areas (P<0.05, AlphaSim-corrected, Figure 1, Table 3), including the left lateral PFC (BA10, BA46, BA47, peak MNI coordinates, x=−51, y=48, z=0, r=−0.78, cluster P<0.001, ReHo mean 0.859±0.037), left medial PFC (BA10, peak MNI coordinates, x=−15, y=63, z=3, r=−0.74, cluster P=0.001, ReHo mean 0.892±0.045), and right lateral orbitofrontal cortex (BA11, BA47, peak MNI coordinates, x=33, y=51, z=−15, r=−0.74, cluster P=0.001, ReHo mean 0.883±0.038). Patients with higher pain scores exhibited lower ReHo values in these prefrontal areas (Figure 1), and vice versa, which suggested that activity of local circuits in the prefrontal areas was modulated by chronic pain. Both pain scores (W₁₆=0.961, P=0.672) and ReHo values (left lateral PFC, W₁₆=0.974, P=0.901; left medial PFC, W₁₆=0.951, P=0.502; right orbital frontal cortex, W₁₆=0.959, P=0.649) conformed to normal distribution, and thus a parametric method (Pearson’s correlation) was used for the statistical analysis.

In order to exclude the possibility that age and sex might have affected these results, we repeated the analysis with the addition of age and sex as covariates of no interest. The results were similar to those of the original analysis. The P-value remained significant (P<0.01, uncorrected), although it did not pass the correction for multiple comparisons. This might have been due to the small sample size of the present study, and thus statistics may be slightly affected by the reduced number of degrees of freedom. Results of this analysis are summarized in Figure S1.

This work was supported by a research fund from the MIND Research Institute, California, to XWD. This work was also supported by the Key Specialist Projects of Shanghai Municipal Commission of Health and Family Planning (ZK2015B01) and the Programs Foundation of Shanghai Municipal Commission of Health and Family Planning (201540114). We are grateful to Dr Yongdi Zhou for his tremendous contributions to the manuscript. We thank all the volunteers for their participation in this study and thank the Shanghai Key Laboratory of Magnetic Resonance, East China Normal University for materials and technical support.