Frequency specific expression of resting state networks in cluster headache

Since in CH the pain is strictly unilateral, we normalized the data according to the headache side: the images of patients having headache on the left side were inverted along the midsagittal axis by the fslswapdim command. To acknowledge that there are non-pain related differences between the hemispheres we repeated the analysis by inverting the images of the patients who had headache on the right side. These results are presented in the main text. In order to confirm that we are not missing any important differences between cluster headache patients and normal controls we repeated the analysis on the raw, non-normalised data.

Methods

The rest of the preprocessing steps and the image analysis was identical to the analysis of the normalised data and detailed in the main text: non-brain structures were stripped from the images, data were motion corrected, smoothed spatially and high-pass filtered in the temporal domain. Functional images were registered to standard brain by a linear and a consecutive non-linear registration.

The resting state networks were identified by independent component analysis. The resting state networks were remapped to the individual functional images space. The mean activity of the network was extracted for all subjects and decomposed into five frequency bands by discrete wavelet decomposition. These band-pass filtered timecourses were then re-regressed to the preprocessed resting fMRI data. The frequency specific spatial maps were compared between groups by GLM approach and nonparametric permutation test was used for statistical comparison.

Results

A significantly increased degree of coactivation was found in the patient group within the right attention network in the 0.04-0.08Hz frequency range in the right superior frontal gyrus (x=14mm, y=34mm, z=44mm, p<0.026 corrected, Figure S1/b). A similar difference was observable in the left superior frontal gyrus in the left attention network (x=-6mm, y=38mm, z=44mm, p<0.0194 corrected Figure S1/a). In the cerebellar network there was a significantly higher degree of coactivation in cluster headache patients in the 0.02-0.04Hz range in the inferior cerebellar vermis (x=2mm, y=-46mm, z=-44mm, p<0.019 corrected, Figure S1/c).

Figure S1. Altered activity in the raw dataset. The alterations of the left (a) and right (b) attention and the cerebellar (c,d) network are depicted on the image. Statistical maps are overlayed on the standard MNI152 brain. The colorbar represents p values.