Participants
MOH patients were recruited from the headache clinic of the Neurological Department of the University Hospital in Bordeaux. They were included if they fulfilled the diagnostic criteria for MOH with prior migraine (ICHD-II 8.2). The exclusion criteria were the following: post-traumatic headaches (ICHD-II 5.1 and 5.2); illness interfering with central nervous system functioning; psychotic disorder or current mood disorder; regular use of a psychotropic medication (antidepressant, benzodiazepine, antipsychotic drugs, but anticonvulsant medications were accepted). Healthy controls (HCs) were recruited among hospital staff members. The two groups were matched to have comparable age and sex distributions. We obtained local ethics committee approval and written informed consent from all subjects before study initiation.
Disease characteristics
MOH patients had to provide information about the age at onset of migraine, the duration of migraine illness, the number of days with headache per month, the type of acute headache medication taken and the mean number of headache medication taken per month.
Neuropsychological evaluation
All subjects filled in several self-administered questionnaires and scales: the Beck Depression Inventory (BDI-13 items [8]), the Pain Catastrophizing Scale (PCS [9]), the State Anxiety Inventory (STAI-state [10]), the Barrat Impulsivity Scale (BIS-11 [11]), and the Medication Dependence Questionnaire for Headache sufferers (MDQ-H [12]). They also underwent an Iowa Gambling Task (IGT [13]) in order to detect putative decision-making impairments.
Magnetic resonance imaging (MRI) acquisitions
Patients were off medication for at least 12 h before the MRI acquisition. MRI scans were obtained using an ACHIEVA 3T scanner (Philips Medical System, Netherlands) with a SENSE 8-channel head coil. Anatomical high resolution MRI volumes were acquired in transverse plan for each subject using a 3D MPRAGE weighted-T1 sequence with the following parameters: TR = 8.2 ms, TE = 3.5 ms, 7-degree flip angle, FOV 256 × 256 mm2 to cover the whole brain, yielding 180 slices, no gap, voxel size 1 × 1 × 1 mm3. Two diffusion-weighted images with opposite polarity, allowing elimination of diffusion imaging gradient cross-terms, were performed using a spin echo single shot EPI sequence with the following parameters: TR = 7646 ms, TE = 60 ms, 90-degree flip angle, FOV 224 × 224 mm2, yielding 60 slices, no gap, voxel size 2 × 2 × 2 mm3. One b0 image was acquired and diffusion gradients were applied in 21 non-collinear directions (b value = 1000 s/mm2). To increase signal-to-noise ratio, the sequence was repeated in two successive runs for each polarity. All acquisitions were aligned on the anterior commissure–posterior commissure plan (AC–PC). For qualitative clinical readings, fluid-attenuated inversion recovery (FLAIR) images were also obtained with the following parameters: TR = 11,000 ms, TE = 140 ms, TI = 2800 ms, FOV 230 × 172 mm2, yielding 24 slices, gap of 1 mm, voxel size 0.72 × 1.20 × 5 mm3. The total scan duration, which also included resting state functional acquisition, was about 45 min.
Volumetric analyses
FreeSurfer 4.5.0 (http://surfer.nmr.mgh.harvard.edu/) is a set of tools for automated cortical and subcortical reconstruction, volumetric segmentation and analysis of images; it parcellates the cortex into gross anatomical regions and produces statistics on volume for each region [14]. Briefly, this processing includes motion correction and averaging of multiple volumetric T1 weighted images, removal of non-brain tissue using a hybrid surface deformation procedure, automated Talairach transformation, segmentation of the subcortical white matter and deep grey matter volumetric structures (including hippocampus, amygdala, caudate, putamen), intensity normalization, tessellation of the grey matter and white matter boundary, automated topology correction, and surface deformation following intensity gradients to optimally place the grey/white and grey/cerebrospinal fluid borders at the location where the greatest shift in intensity defines the transition to the other tissue class. All stages of the stream are fully described by Fischl et al. [14].
To extract reliable volume estimates, Freesurfer’s semi-automatic anatomical processing was executed on the T1-weighted images of the two groups of subjects. The volume measures were therefore subject-specific. Grey matter was segmented into 148 regions, based on the Destrieux atlas [15]. Subcortical structures and ventricular system were segmented into 40 regions, based upon the existence of an atlas containing probabilistic information on the location of structures [16]. Finally, visual verifications were performed on every subject. In order to correct for inter-subject variability, each volume of the segmented regions was divided by the total intracranial volume for each subject.
Tractography seeds
The regions of interest (ROIs) were manually defined (co-author M.E.) using two atlases: the Atlas of Human brain, 3rd edition, 2007 and the atlas of Duvernoy [17]. We chose manual delineation and not FreeSurfer segmented label because previous study has shown that boundaries of the Freesurfer parcellation does not perfectly match with anatomical boundaries of this region specifically [18]. As seed masks for tractography analysis, the ROIs included the bilateral hippocampus, drawn on individual T2 scans (b0 diffusion unweighted image) (see Fig. 1 for an illustrative example of the seed mask).
The hippocampi were traced in the coronal plane with correction in the axial and sagittal planes. First, ROIs were delineated in the anterior part of the hippocampus (head) i.e., the posterior part of the amygdala to the tip of the tail of the hippocampus at the atrium. The temporal horn of the lateral ventricle (superior and lateral), the uncus of the apex then the Fimbria delimited the hippocampus body. A thin band of white matter was included along the Ammon Horn (inferior) and reached the Subiculum, which was included in its proximal part (median). Finally, the Crus Fornicis (superior), the lateral ventricle (lateral) and the Atrium defined the tail of the hippocampus.
Then, corrections in axial and sagittal planes were made on areas of white matter appearing in coronal plane. Finally, delimitation of the amygdala and boundaries with the pons were done on same planes. Regions of interest were enlarged with a two-voxel extension (4 mm) laterally to include white matter to ease tractography.
Tractography
The fiber components of bundles emerging from ROIs were reconstructed using a voxel-by-voxel regularized streamline algorithm [19], which resembles diffusion tensor deflection [20, 21] and which is implemented in BrainVISA software (http://brainvisa.info/). This strategy, similar to diffusion tensor deflection, overcomes simple crossing configurations. At each voxel, two parameters were used to get the direction of the resulting tract: the tensor biggest eigenvectors directions with a weight of α (α representing local anisotropy) and inertia (that is, the incident direction of the tract) with a weight of 1 − α. The algorithm stepped forwards a distance of half the voxel size along this direction and did the computation again. The propagation mask excluded voxels likely to belong to grey matter (Fractional anisotropy FA < 0.2), and algorithm stopped for a maximum curvature angle adjusted for each bundle to 45° or when the tract length exceeds 200 mm. In addition to the number of reconstructed fibers, mean FA and mean ADC (Apparent Diffusion Coefficient) along the reconstructed fiber tracts were calculated.
In this article, the term ‘tracked fibers’ refers to the number of fibers generated by the algorithm in fiber bundles connecting predefined brain regions, rather than the actual number of fibers in anatomical fiber bundles (for further explanation, see [22]).
In order to correct for inter-subject hippocampal volume variability, the number of “fibers tracked” was divided by the hippocampal volume.
Statistical analyses
Data were analysed using the Statistical Package for Social Sciences IBM SPSS 20.0. (IBM Corp., Armonk, NY). The significance level was set at p < 0.05. We studied variables distribution with the Kolmogorov–Smirnov test and brain regional volumes differing between groups using t tests for independent samples; no correction for multiple comparisons was performed.
Within the group of MOH, we calculated Spearman’s nonparametric correlations between brain regional volumes which differed between groups, disease characteristics and neuropsychological evaluation.
We calculated group differences for the ROI volume, and with respect to tractography results, that is number of reconstructed fibers as well as mean FA and mean ADC along the respective reconstructed fiber tract. Model check was performed using residual QQ plots and detection of outliers using leaf plots. If outliers were detected, they were removed from analyses.