Subjects and clinical data
Seventy-five subjects, including 45 patients with chronic bilateral tinnitus and 50 non-tinnitus controls (all right-handed and completed at least 8 years of education), were recruited through community health screening and newspaper advertisements and matched for sex, age, and education. According to the International Classification of Headache Disorders, Third Edition (beta version) (ICHD−3 beta) [19] as well as the headache symptom of each subject, tinnitus patients were divided into two groups (20 patients with headache and 25 patients without headache). The Iowa version of the Tinnitus Handicap Questionnaires (THQ) [20] as well as a pure tone audiometry (PTA) examination was used to assess the tinnitus severity, tinnitus distress, and the hearing threshold. Any participants who had hearing loss (defined as thresholds≥25 dB HL) at the frequencies of 0.25 kHz, 0.5 kHz, 1 kHz, 2 kHz, 4 kHz, and 8 kHz were excluded from the current study. There were no significant differences of auditory thresholds between tinnitus patients and non-tinnitus controls (Figure 1). Participants were excluded if they suffered from pulsatile tinnitus, hyperacusis or Meniere’s diseases, or if they had a past history of alcoholism, stroke, migraine, brain injury, anemia, Alzheimer’s disease, Parkinson's disease, epilepsy, major depression or other neurological or psychiatric illness, MRI contraindications or severe visual loss, thyroid dysfunction, cancer, severe heart diseases and damaged liver/kidney function.
According to the Self-Rating Depression Scale (SDS) and the Self-Rating Anxiety Scale (SAS) (overall scores < 50, respectively), none of the participants had depression or anxiety. The pain intensity of headache was measured by the visual analogue scale (VAS) and degree of headache was measured by the Headache Impact Test-6 (HIT-6). Demographics and clinical characteristic data of the chronic tinnitus patients and non-tinnitus controls were summarized in Table 1.
MRI data acquisition
All participants were scanned using a 3.0 T MRI scanner (Ingenia, Philips Medical Systems, Netherlands). Foam padding and earplugs were used to reduce the head motion and scanner noise. The participants were instructed to rest quietly with their eyes closed and avoiding either falling asleep or making sudden head motions, and to not think of anything in particular during MRI scan. High resolution three-dimensional turbo fast echo (3D-TFE) was acquired using the parameters as follows: repetition time (TR) = 8.1 ms, echo time (TE) = 3.7 ms, thickness = 1 mm, slices = 170, gap = 0 mm, flip angle (FA) = 8°, field of view (FOV) = 256 × 256 mm2, and acquisition matrix = 256 × 256. The structural sequence took 5 minutes and 29 seconds. ASL images were obtained with a pseudo-continuous ASL (pcASL) sequence with a 2D fast spin-echo acquisition and background suppression using the parameters as follows: TR = 4000 ms, TE = 11 ms, slice thickness = 4 mm, label duration=1650ms; post-label delay = 1600 ms, FA = 90°, FOV = 220×220 mm2, slices thickness = 4 mm, gap = 0.4 mm, reconstruction matrix = 672. The ASL sequence took 4 minutes and 18 seconds.
Imaging data processing
A voxel-based morphometry (VBM) approach was performed to estimate whole brain volumes using the VBM8 toolbox (http://dbm.neuro.uni-jena.de/vbm). DARTEL was used to improve inter-subject registration of the structural images. Briefly, cerebral tissues were segmented into GM, white matter (WM), and cerebrospinal fluid by a unified segmentation algorithm [21]. Then, resulting GM and WM images were normalized to the MNI template, followed by smoothing using an 8-mm full width at half maximum (FWHM) Gaussian kernel. Finally, the resulting voxel-wise GM volume maps were entered as covariates in the ASL data analysis.
The ASL data were preprocessed to generate CBF maps using the ASL Perfusion MRI Signal Processing Toolbox (ASLtbx), which is based on SPM12 (http://www.fil.ion.ucl.ac.uk/spm/) [22]. All images were first rearranged and adjusted to correct head movement. Next, a nonlinear transformation was performed on the CBF images of healthy controls, which were co-registered with the PET- perfusion template in Montreal Neurological Institute (MNI) space. The MNI-standard CBF template was defined as the average co-registered CBF images of healthy controls. The CBF images of all participants were then co-registered to the MNI-standard CBF template. Every co-registered CBF was removed from the non-brain tissue. Then a spatial smoothing with an isotropic Gaussian at FWHM of 6 mm3 was followed. Finally, normalization was performed by dividing the cerebral blood flow per voxel by the average cerebral blood flow across the entire brain [23]. None of the participants was excluded from the study due to head movement exceeding 2.0 mm of maximum translation in any of the x, y, and z directions or 2.0°of the maximum rotation around the three axes.
Statistical analyses
Clinical measures were analyzed using Statistical Package for Social Sciences (SPSS) statistics software package version 20.0 (IBM Corp., Armonk, NY, USA). The statistical significance level was set at p<0.05, two-tailed. One-way analysis of variance (ANOVA) was used to calculate the difference among the three groups followed by a post hoc test (t-test for means and χ2-test for proportions) between tinnitus patients with headache and patients without headache.
A one-way analysis of variance (ANOVA) was then performed to determine between-group differences in brain volumes, with age, sex, and education as the nuisance covariates. Between-group differences in CBF were also calculated via one-way ANOVA in SPM12 with age, gender, education level and GM volume as the nuisance covariates. Significant thresholds were corrected using false discovery rate (FDR) criterion and set at p<0.01. A full-factorial model was utilized to detect potential interaction effects between tinnitus and headache on CBF differences. Significant thresholds were corrected using cluster-level family-wise error (FWE), and the threshold was set at p < 0.01.
The relationships between aberrant CBF and each clinical characteristic were further investigated. Firstly, regions showing significant differences between groups were extracted. Then the mean z-values of aberrant CBF region mask were calculated within every subject. Pearson correlation analysis between the mean z-values and each clinical characteristic were performed using SPSS software. Partial correlations were calculated with age, sex, education, GM volume, and average hearing thresholds as the nuisance covariates. P < 0.05 was considered statistically significant.