Participants
Patients with idiopathic CD, and HCs were recruited from our outpatient movement disorder clinic between September 2021 and July 2023. All patients met the following inclusion criteria: (i) age 19–75 years; and (ii) a diagnosis of CD was established according to the published criteria by an experienced neurologist (G.L.) [18]. Exclusion criteria were as follows: (i) known causes of dystonia or a family history of movement disorders; (ii) had dystonia involving other body sites in addition to neck muscles; (iii) reported evidence of Parkinson’s disease, stroke, traumatic brain injury, Alzheimer’s disease, and epilepsy; (iv) had a family history of movement disorders as well as a history of exposure to antipsychotic drugs before the onset of dystonia; (v) had any conditions that contradicted with cerebral MRI; (vi) received botulinum toxin (BoNT) injections within 3 months and/or oral medications for approximately 24 h before MRI scans. HCs were recruited using the same exclusion criteria. Written informed consent was obtained from all the participants and the study was approved by the Ethical Committee of the First Affiliated Hospital of Sun Yat-sen University ([2020]323).
Clinical assessment
Demographics and clinical characteristics, including patients’ age, sex, duration of disease, and number of BoNT injections were collected from all patients via face-to-face interviews before MRI scanning. The Toronto Western Spasmodic Torticollis Rating Scale (TWSTRS), Global Dystonia Rating Scale (GDS), and Cervical Dystonia Impact Profile-58 (CDIP-58) were used to assess the severity, disability, and effects of CD on the quality of life [19–21]. Non-motor symptoms, including anxiety, depression, and cognition situation were assessed using the Hamilton Anxiety Scale (HAMA) [22], Hamilton Depression Scale (HAMD) [23], and Minimum Mental State Examination (MMSE) [24], respectively.
Data acquisition
MRI data for each participant were collected using a 3T scanner (Tim Trio; Siemens, Erlangen, Germany). DTI data were acquired using a single-shot echo-planar imaging sequence (repetition time, 7000 ms; echo time, 91 ms; flip angle, 90°; acquisition matrix, 128 × 128; field of view, 256 × 256 mm2; voxel size, 2 × 2 × 3 mm3; 50 axial slices). Diffusion weighting was isotopically distributed in 64 directions using a b value of 1000 s/mm2. Moreover, three dimensional T1-weighted images (repetition time = 2530 ms, echo time = 4.45 ms, inversion time = 1100 ms, flip angle = 7°, matrix dimensions = 256 mm × 256 mm, voxel size = 1 × 1 × 1 mm3, and 192 slices) were obtained to improve registration to the standard space.
Image preprocessing
All DTI data were analyzed using the PANDA toolbox (Pipeline for Analyzing Brain Diffusion Images toolkit, https://www.nitrc.org/projects/panda/) with structural MRI of the brain [25]. First, the brain mask was estimated for each participant from the b0 image, and the non-brain tissues were removed. Subsequently, the data were corrected for head motion and eddy current distortion by applying the affine registration of each diffusion-weighted image to the b0 image. The diffusion tensor elements were then estimated, and the fractional anisotropy (FA) was calculated for each voxel. The generated FA images were registered to the MNI 152 standard space using nonlinear registration. Subsequently, the whole-brain fiber tractography was performed by a deterministic tracking algorithm using the Diffusion Toolkit (http://trackvis.org) and TrackVis software (http://trackvis.org/) [26]. All tracts were computed by seeding each voxel with a FA > 0.2. Tractography was terminated if it turned at an angle exceeding 45° or reached a voxel with a FA of less than 0.2.
Brain network construction
Adopting the approach used in our previous study (Guo et al., 2021)[15], network nodes were defined using the BNA-246 atlas (http://atlas.brainnetome.org/). This fine-grained parcellation atlas includes more detailed information on both functional and anatomical connections, which could help to describe connectivity in the brain global network characteristics more accurately [27]. We set the fraction of streamline (FSe) values as the edge weights of the network [28]. Finally, a symmetric FSe-weighted (246 × 246) matrix, representing the WM structural network, was generated for each participant (Fig. 1).
Global network measures
On the global level, the integration (e.g., global efficiency [Eglob] and shortest path length [Lp]), segregation (local efficiency [Eloc], cluster coefficient [Cp], and modularity [Q]), and resilience (assortativity [r]) measures [29] of WM anatomical networks were computed based on the FSe-weighted (246 × 246) matrix for each participant by using the Gretna toolbox (https://helab.bnu.edu.cn/gretna/) [30], which was implemented using the MATLAB (R2018b) platform (https://www.mathworks.com/). In addition, network strength (Sp) and small-world properties were assessed to characterize the topological organization of the networks. Eglob can indicate the efficiency of information transference across a network, while Eloc indicates how well a node exchanges information with its neighboring nodes [31]. A smaller Lp indicates the faster information transfer to the entire brain region, and Cp quantifies the prevalence of clustered connectivity around individual nodes [32]. Additionally, assortativity, known as degree correlation, is a measure of the correlation between nodal degree and mean degree of its nearest neighbors, which is related to the more vulnerable network being attacked by lower assortativity [33–35]. Positive assortativity values indicate that nodes may be connected to other nodes of the same degree, and high-degree nodes or hubs of the network are likely to be connected [29]. A small-world network is defined as γ > 1 and λ ≈ 1. These two measurements can be summarized into a simple quantitative metric, small worldness, σ = γ/λ > 1 [36].
Regional network measures
We calculated the nodal efficiency (Enodal) and two metrics of nodal centrality, betweenness centrality (BC) and degree centrality (DC), for all 246 regions [37]. Enodal represents the importance of a given node in conveying information within a network. BC is the fraction of all shortest paths in the network that passes through a given node. The DC, the number of links connected to a node, is one of the most common centrality measures.
Statistical analyses
Differences in age between patients with CD and HCs were assessed using two sample t-tests or Mann-Whitney U tests after Shapiro-Wilk normality testing. Sex distributions were compared using the Pearson χ2 test. All analyses were performed using SPSS (version 27.0; IBM, Armonk, NY, USA).
Between-group differences in network parameters were determined using two sample t-test, with HAMA and HAMD scores as covariates. In addition, the relationships between the abnormal network graph theoretical metrics and clinical features, including TWSTRS total scores, TWSTRS subscales, and disease duration, were assessed using partial correlation analyses after adjusting for age, sex, HAMA, and HAMD scores. P < 0.05 was set to evaluate statistical significance.
Reproducibility analyses
Referring to the previous study [15], we evaluated the potential effects of different parcellation schemes employing similar network analyses with an additional Anatomical Automatic Labeling atlas with 90 brain regions (AAL-90; https://www.gin.cnrs.fr/en/tools/aal/) in patients with CD [38].