Patients with peripheral neuropathy and healthy volunteers will be recruited in this study. Peripheral neuropathy is defined according to the neuropathic symptoms and signs. Informed consent will be approved by the Ethical Committee of the National Taiwan University Hospital and obtained from each subject.

To assess the severity of different neuropathic symptoms, such as spontaneous ongoing and paroxysmal pain, evoked pain, paraesthesia, and dysaesthesia, patients with neuropathic pain will fill out the Neuropathic Pain Symptom Inventory. Each subject will receive detailed sensory examination to evaluate the integrity of sensory fibers. To measure thresholds of thermal and vibratory sensations, we will perform quantitative sensory testing by the method of level using a Thermal Sensory Analyser and Vibratory Sensory Analyser (Medoc Advanced Medical System, Minneapolis, MN, USA) following an established protocol. We will use a contact heat stimulator to deliver thermal stimulation. Noxious and innocuous heat temperatures will be applied within the right foot dorsum. Several pretests will be applied before CHEP recording to eliminate expectation effects. To avoid sensitization and desensitization, low intensity stimuli will precede high intensity stimuli at each block.

Functional magnetic resonance imaging (fMRI) will be performed on a 3-T MR machine (Sonata; Siemens, Erlangen, Germany). A high resolution T1 weighted scan of the entire brain in trans-axial orientation will be obtained to provide anatomical information for the superimposed functional activation maps. Echo Planar Imaging will be used for the acquisition of the functional data. Each imaging session will be consisted of one high-resolution anatomical scan and three functional scanning runs, with 5-min intersession interval. During the scanning, several thermal stimuli will be applied by CHEP stimulator to the right dorsal foot. To avoid sensitization, the stimulation site will be changed slightly after each stimulus. After 12-s stimulation, the temperature will be cooling, with a subsequent 36-s interstimulus interval. After each fMRI session, subjects will be asked to rate the intensity and unpleasantness of received pain stimulus. The average rating values will be indicated after the scan on a computer driven visual analogue scale ranging from 0 to 10 (0, no pain; 1, slight intense; 2, mild intense; 3, moderate intense; 4, slight pain; 5, mild pain; 6, moderate pain; 7, moderate-strong pain; 8, strong pain; 9, severe pain; 10, unbearable pain), and the intensity and unpleasantness of received pain will be assessed using the Short-Form McGill Pain Questionnaire.

All data will be processed using the Statistical Parametric Mapping software (SPM2). fMRI data series will be realigned and resliced with sinc interpolation to correct for motion artifacts. Scans with sudden head movements of more than 2 mm will be omitted. To enable intersubject analysis, the functional data will be coregistered to the anatomical scan and transformed into a reference space according to the MNI template of SPM2 by normalization using sinc interpolation. The resampled voxel volume of the normalized images is 2 x 2 x 2 mm. Subsequently, data will be smoothed with an isotropic Gaussian kernel of 8 mm full-width at half maximum to reduce high frequency noise and to account for anatomical variances. Condition-specific effects will be estimated with the general linear model using a boxcar approach convolved with the hemodynamic response function. High pass filtering will remove low frequency noise and low pass filtering will account for serial autocorrelations of the data.

We will analyze the data on an individual (subject per subject) basis and across subjects (group analysis) using a cross-subjects variance (random effect analysis). Data from each stimulation will be pooled for group statistical comparisons. A single design matrix, including 3 sessions of all subjects, will be generated due to the limited number of experiments representing a fixed-effects model analysis. Statistical parametric maps will be generated as t-contrasts and corrected for multiple comparisons according to the random field theory with P < 0.05. The threshold for the Z maps is 3.09 (P < 0.001) for individual subject analysis. Significant clusters have to show a minimum extension volume of 10 voxels.