Distinctive alterations in the functional anatomy of the cerebral cortex in pain-sensitized osteoarthritis and fibromyalgia patients

Background Pain-sensitized osteoarthritis and fibromyalgia patients characteristically show nociceptive system augmented responsiveness as a common feature. However, sensitization can be originally related to the peripheral injury in osteoarthritis patients, whereas pain and bodily discomfort spontaneously occur in fibromyalgia with no apparent origin. We investigated the distinct functional repercussion of pain sensitization in the cerebral cortex in both conditions. Methods Thirty-one pain-sensitized knee osteoarthritis patients and 38 fibromyalgia patients were compared with matched control groups. And new samples of 34 sensitized knee osteoarthritis and 63 fibromyalgia patients were used to directly compare each condition. A combined measure of local functional connectivity was estimated to map functional alterations in the cerebral cortex at rest. Results In osteoarthritis, weaker local connectivity was identified in the insula, which is a cortical area processing important aspects of the brain response to painful stimulation. In contrast, fibromyalgia patients showed weaker connectivity in the sensorimotor cortex extensively affecting the cortical representation of the body. Conclusions In osteoarthritis, weaker insular cortex connectivity is compatible with reduced neural activity during metabolic recovery after repeated activation. In the fibromyalgia neurophysiological context, weaker connectivity may better express both reduced neural activity and increased excitability, particularly affecting the sensorimotor cortex in patients with spontaneous body pain. Such a combination is compatible with a central gain enhancement mechanism, where low sensory tolerance results from the over-amplification of central sensory reception to compensate a presumably weak sensory input. We propose that deficient proprioception could be a factor contributing to weak sensory input. Supplementary Information The online version contains supplementary material available at 10.1186/s13075-022-02942-3.


Iso-Distant Average Correlation (IDAC) maps
A novel mapping was used to characterize the functional structure of the cerebral cortex based on Iso-Distant Average Correlation (IDAC) measures. Essentially, IDAC mapping expands well-established MRI measures of local functional connectivity [6][7][8] by combining the connectivity maps of varying distances. Composite IDAC maps may uniquely inform the connectivity-related specialization of the cerebral cortex as local connectivity is distancespecific to a large extent and proved to discriminate well between major classical anatomofunctional cortical areas [2,5].

Image processing
Imaging data were processed using MATLAB version 2016a (The MathWorks Inc, Natick, Mass) and Statistical Parametric Mapping software (SPM12; The Wellcome Department of Imaging Neuroscience, London).
Anatomical and functional images were visually inspected to detect possible acquisition artifacts. Functional MRI images were slice-time corrected, realigned and then smoothed by convolving the image with a 4x4x4mm3 full width at half maximum (FWHM) Gaussian kernel.
The resulting realignment parameters were used for scrubbing, namely, discarding motionaffected volumes [3]. For each subject, mean inter-frame motion measurements [4] served as an index of data quality to flag volumes of suspect quality across the run. At time points with mean inter-frame motion > 0.3 mm, the corresponding volume, the immediately preceding and the succeeding two volumes were all discarded. Using this procedure, a mean (± SD) of 3.4 (± 10.2) volumes from the total of 180 volumes included in the fMRI sequence were removed in the Osteoarthritis Sample 1; mean 2.7 (± 6.2) in the corresponding control group; mean 3.6 (± 5.9) in the Fibromyalgia Sample 1; mean 6.2 (± 12.8) in the corresponding control; mean 5.3 (± 10.6) in the Osteoarthritis Sample 2; and mean 0.9 (± 2.5) in the Fibromyalgia Sample 2.
Image volumes were then co-registered to their anatomical images with an affine transformation. A warping matrix was also estimated for every subject to match a group template created from the 3D anatomical individual acquisitions and then to the Montreal Neurological Institute (MNI) space using DARTEL normalization [1]. Image volumes were re-sliced to 3x3x3 mm. Estimated DARTEL normalizations to the MNI space were applied to the IDAC results to enable group inferences.
IDAC computations (see below) were conducted in a gray matter mask split into left and right hemispheres, so that no adjacent voxels from the medial regions of one hemisphere would be locally associated with those from the other hemisphere. The two hemispheres were brought back together once the IDAC values had been calculated. The left and right hemisphere gray matter masks were obtained by setting a threshold of p>0.4 on the gray matter probability maps obtained from the DARTEL group template. As IDAC value estimations were carried out in every subject's native space, the template masks were back-transformed with the inverse Whole-cortex IDAC maps were generated by estimating the average temporal correlation of each voxel with all its neighboring voxels placed at increasingly separated Euclidean iso-distant intervals. IDAC was computed in native space separately for each hemisphere after realignment and smoothing. Three IDAC maps were obtained at distance intervals 5-10mm, 15-20mm and 25-30mm.
Multi-distance IDAC color maps were obtained from the overlay of the three IDAC maps using an RGB color codification (see Figures). RGB color channels enabled the display of three values simultaneously. RED corresponding to the results from 5-10mm IDAC map analyses, GREEN from 15-20mm and BLUE from 25-30mm. The overlapping of these primary colors produces a full range of secondary colors.

Definition of Iso-Distant Average Correlation (IDAC)
We defined the concept of "Iso-Distant Average Correlation" (IDAC) to describe the pattern of correlation decay in the close vicinity of a voxel [2]. IDACi(h) was consequently defined as the average temporal correlation of voxel i with all the voxels located at a given Euclidean distance interval h. Functional MRI data sets being a discrete sample, any distance interval h must be necessarily transformed into a discrete iso-distant interval Hk=(hk, hk+1), with hk being a set of successively increasing distances covering the whole vicinity of a given voxel (see Figure).
The set of iso-distant intervals Hk were selected so that temporal correlations were mainly positive, decreased monotonically and in which horizontal axon collaterals were considered likely to form local networks. For the present study, we defined 3 iso-distant intervals: 5-10mm, 15-20mm and 25-30mm, with constant thicknesses but increasing number of voxels.
We first computed a correlation matrix C of Pearson coefficients comparing the functional MRI signal time course of all the voxels in our study mask with each other's.
where M is the length of the functional MRI signal time series and i and j index all the voxels entering our study mask. We then transformed the Pearson correlation matrix C into a Gaussian distributed z-score correlation matrix Z by applying a Fisher transform.  interval Hk,i. Note that, for a given distance interval k, the number of voxels within the concentric iso-distant interval Nk,i is not necessarily the same for every voxel i due to the edge effects of the study mask. Figure. fMRI Temporal correlations between one voxel ("seed") and its neighboring peripheries present a characteristic decreasing spatial gradient. LEFT: Fisher-transformed z-scores of a correlation map with a "seed" voxel in the visual area from a single subject. Voxel resolution is 3x3x3mm and results are constrained to distance intervals hk<30mm and within the subject's native gray-matter mask (blue shade). RIGHT: Six different Iso-distant intervals as they are used to calculate different IDAC values in our study.
Supplementary Figure 2. Negative correlation between age and IDAC values (short distance, 5-10 mm) in knee osteoarthritis patients at a lower threshold (clusters at p< 0.01 surviving whole-brain family-wise error (FWE) correction (p< 0.05)). This analysis indicates that, although the effect was marginal in osteoarthritis, it was also in the direction of age-related local connectivity weakening in auditory cortex in osteoarthritis.

Supplementary Tables
Supplementary Table 1