Increased connectivity of pain matrix in chronic migraine: a resting-state functional MRI study

Abstract

Objective: To investigate the whole-brain resting-state functional connectivity in patients with chronic migraine (CM) using a data-driven method.

Methods: We prospectively recruited patients with either episodic migraine (EM) or CM aged 18-60 years who visited the headache clinic of the Samsung Medical Center from July 2016 to December 2017. All patients underwent 3 T MRI using an identical scanner. Patients were considered interictal if they did not have a migraine headache at the day and ± 1 days of functional MRI acquisition. Using the group-independent component analysis (ICA), connectivity analysis with a weighted and undirected network model was performed. The between-group differences in degree centrality (DC) values were assessed using 5000 permutation tests corrected with false discovery rate (FDR).

Results: A total of 62 patients (44 EM and 18 CM) were enrolled in this study. Among the seven functionally interpretable spatially independent components (ICs) identified, only one IC, interpreted as the pain matrix, showed a significant between-group difference in DC (CM > EM, p = 0.046). This association remained significant after adjustment for age, sex, migraine with aura (MWA), allodynia, depression, and anxiety (p = 0.038). The pain matrix was functionally correlated with the hypothalamus (p = 0.040, EM > CM) and dorsal raphe nucleus (p = 0.039, CM > EM) with different levels of strength in EM and CM.

Conclusion: CM patients have a stronger connectivity in the pain matrix than do EM patients. Functional alteration of the pain network might play a role in migraine chronification.

Keywords: Chronic migraine; Functional MRI; Migraine; Neuroimaging; Pain matrix.

Fig. 1

Region of interest segmentation results. The hypothalamus, dorsal raphe nuclei (DRN), and periaqueductal gray (PAG) were segmented on a three-dimensional brain atlas

Fig. 2

Resting-state networks identified using independent component analysis. Nine automatically generated independent components (ICs). ICs 8 and 9 were considered as noise components and were therefore excluded. ICs 1 to 7 are functionally interpretable ICs

Fig. 3

Pain matrix. The functional network (IC 3) included the dorsolateral prefrontal cortex, anterior insular cortex, anterior cingulate cortex, thalamus, and precuneus, suggesting a pain matrix

Fig. 4

Group comparison among identified resting-state networks. DC values were compared between EM and CM by using permutation tests with FDR. Each bar and error bar represent the mean and standard errors of mean, respectively

Fig. 5

Functional connectivity of IC 3 (pain matrix) with the hypothalamus, DRN, and PAG. a Functional connectivity of IC 3 with the hypothalamus, dorsal raphe nuclei (DRN), and periaqueductal gray (PAG) are illustrated. The mean (SD) of edge values are shown and the width of the lines indicate the magnitude of edge weights. b Between-group analysis of connectivity between IC 3 and each region are summarized. Patients with CM showed a stronger connectivity between IC 3 and the hypothalamus (FDR-corrected p = 0.0399), whereas the connectivity between the DRN and IC 3 was stronger in patients with EM (FDR-corrected p = 0.0390). No between-group difference in the connectivity between IC 3 and the PAG were noted (FDR-corrected p = 0.2738)