We studied 70 children: 53 of them were included in this study, while 17 children were excluded due to either severe head movement (n = 8) or inability to complete the magnetic resonance imaging (MRI) scan (n = 9). The study includes 16 TSpure (age: 9.7 ± 2.1, M/F: 15/1), 14 TS + OCD (age: 10.2 ± 2.1, M/F: 10/4), 11 OCD (age: 10.7 ± 2.5, M/F: 7/4), and 12 controls (age: 10 ± 1.2, M/F: 3/9) with episodic tension headache who were headache-free during the MRI scan. Only those controls were enrolled in the study who had no current or prior history of any neurological or psychiatric disorder including tics and OCD. All subjects were enlisted from the child and adolescent neuropsychiatry outpatient clinic at the Department of Human Neurosciences, Sapienza University of Rome, Italy. Inclusion criteria were (a) drug-naivety; (b) right-handedness; and (c) normal cognitive profile (IQ ≥ 70). Exclusion criteria were (a) attention-deficit hyperactivity disorder, autism spectrum disorder, and any axis I psychiatric disorder comorbidity such as schizophrenia, schizoaffective disorder, bipolar disorder, major depression disorder, and eating disorders; (b) prior behavioral treatment; or (c) contraindications to MRI.
Cognitive evaluation for all participants was assessed by means of the Wechsler intelligence scale for children III (WISC-III) full scale. Diagnosis was made according to DSM-5 criteria  by a neuropsychiatrist experienced in assessing pediatric TS, OCD, and related comorbidities. Tic and OCD symptom severity was assessed using the Yale global tic severity scale (YGTSS) total tic score (TTS) (range, 0–50) (YGTSS-TTS: range 0–50, without impairment score) and the children’s Yale-Brown obsessive–compulsive scale (CYBOCS), respectively. For clinical demographics of the participants refer to Table 1. The presence of other developmental disorders including ADHD or psychiatric disorders other than OCD was ruled out by means of the K-SADS-PL parental interview administered to both parents . This study was approved by the institutional review board of Sapienza University of Rome. Written informed consent was obtained from all parents/guardians in accordance with the Declaration of Helsinki.
After clinical evaluation, all subjects underwent a 3 T MRI scan (Magnetic Verio; Siemens, Erlangen, Germany) using a standardized protocol and a 12-channel head coil designed for parallel imaging (GRAPPA, generalized autocalibrating partial parallel acquisition). Subjects were scanned in a supine head-first position with symmetrically placed cushions to minimize head motion. The MRI protocol included the following sequences: (a) high-resolution 3D, T1-weighted (3DT1) MPRAGE: repetition time (TR) = 1900 ms, echo time (TE) = 2.9 ms, flip angle = 9°, field of view (FOV) = 260 mm2, matrix = 256 × 256, 176 sagittal slices 1 mm thick, no gap; (b) diffusion-tensor imaging (DTI, single-shot echo-planar spin-echo sequence, with one b = 0 and 30 gradient directions, b = 0 and 1000 s/mm2, TR = 12,200 ms, TE = 94 ms, FOV = 192 mm, matrix = 96 × 96, 72 axial 2-mm thick slices, no gap); and (c) resting-state functional magnetic resonance imaging (rs-fMRI): TR = 3000 ms, TE = 30 ms, flip angle = 89°, 64 × 64 matrix, 50 contiguous axial slices 3 mm thick, 140 vol, acquisition time = 7 min. During the MRI scan, subjects were asked to lie down, close their eyes, and remain awake and relaxed.
Images were analyzed via FMRIB’s software library (FSL) version 6.0.1. 3DT1 images were brain extracted using the brain extraction toolbox (BET) and segmented into GM, WM, and cerebrospinal fluid (CSF). An age-specific pediatric template was created via the cerebromatic toolbox , with age, sex, and scanner strength as covariates.
Lobular Volume Analysis
Cerebellar lobular volumes were calculated using the SUIT toolbox . The procedure involved cropping and isolating the cerebellum from the 3DT1 anatomical images for each subject. Each cropped image was subsequently normalized into SUIT space using generated flowfield and affine transformations. Lastly, the probabilistic cerebellar atlas was resliced back into the individual subject space, resulting in GM measurements of 13 bilateral regions of the cerebellum (lobules I–IV, V, VI, crus I, II, VIIb, VIIIa, VIIIb, IX, X, dentate, interposed nucleus, and fastigial nucleus) and 8 vermis regions. Values of the extracted cerebellar regions were normalized to the total intracranial volume to reduce head size variability.
Tract-based spatial statistics (TBSS)  was employed to evaluate WM integrity. Head movement- and eddy-corrected images were fitted to the tensor model at each voxel to calculate fractional anisotropy (FA) and mean diffusivity (MD) maps using DTIFIT. Since our subjects were children, the FA target image was unsuitable. Hence, the most representative target image (i.e., the one that required the least amount of warping to match every other subject) across the four groups was identified and all subject-specific FA and MD maps were non-linearly aligned to this target image. Lastly, a WM skeleton was generated from the mean FA image by thresholding at 0.2 to exclude GM or CSF. Automatic tract-specific quantification using the John Hopkin’s University (JHU) WM tractography atlas was performed to identify the inferior cerebellar peduncles (ICP), middle cerebellar peduncles (MCP), and superior cerebellar peduncles (SCP), which were used as masks to restrict DTI analysis. Mean FA and MD values were then computed.
Functional MRI Analysis
The functional MRI expert analysis tool (FEAT) was used to preprocess rs-fMRI data. After discarding the first 3 volumes, rs-fMRI data were slice-time corrected, motion parameters estimated via MCFLIRT, spatially smoothed (Gaussian kernel of full-width half maximum = 8 mm), and high pass temporal filtered (100.0 s). Next, the rs-fMRI data were subjected to ICA-based noise reduction strategy called ICA-AROMA (automatic removal of motion artifacts) . This method uses an automatic classifier that categorizes each component as either BOLD signal or artifact, based on its high frequency content > 35%, correlation with realignment parameters (RP) derived from MCFLIRT, edge and CSF fractions. Additionally, the mean time courses of WM and CSF were calculated from the rs-fMRI images and regressed out via fsl_glm. The rs-fMRI data were then subjected to a bandpass filter at (0.01–0.09) Hz and finally normalized onto the customized T1 template space.
For seed-based analysis, a bilateral spherical seed of the DN (4-mm radius) was created using the coordinates of the left and right dentate (− 18, − 58, − 34 and 18, − 56, − 34, respectively) in MNI space, consistent with our previous work [26, 27]. The dentate seed was then affine-transformed to the native space of each subject and its anatomical location was carefully checked on functional images. The mean time series of the bilateral dentate seed were computed for each subject and inserted into a general linear model (GLM) to generate seed-based correlation maps.
The Kruskal–Wallis test and post hoc Mann–Whitney U test were performed to assess between-group differences with respect to age, followed by the chi-square test to check for inter-group differences with respect to sex. Differences in clinical scores between groups were analyzed via Mann–Whitney U test. Statistical Package for the Social Sciences (SPSS-25.0) was used to compute statistical analyses. Results were Bonferroni corrected at p < 0.05.
Lobular Volume Analysis
Differences in normalized cerebellar GM volumes between study cohorts were calculated using MANCOVA followed by a post hoc two-sample t-test. Lastly, Pearson’s correlation was computed between the cerebellar lobules that significantly differed from controls and clinical scores. In all analyses, age and gender were included as covariates. Results were Bonferroni corrected at p < 0.05.
To compute between-group differences in FA and MD in the three cerebellar peduncles, a GLM was constructed with age and gender as covariates. An ANOVA followed by a two-sample t-test was run to investigate inter-group differences via a non-parametric approach (applying, 5000 permutations) using Randomise tool  of FSL. Results were false discovery rate (FDR)-corrected with a p value < 0.05. For the correlation analysis, the mean values of significantly altered FA and MD between patients and controls were extracted. Pearson’s correlation was computed between mean FA and MD values and severity scores, and results were Bonferroni corrected at p < 0.05.
Dentate Nucleus Functional Connectivity (DNFC)
To evaluate inter-group differences in terms of DNFC, ANOVA was performed non-parametrically via Randomise tool  (n = 5000) using a GLM with age and sex as covariates to map brain areas that significantly differed between groups. The ANOVA-derived map was binarized to build a mask, at p < 0.05 after FDR correction. Furthermore, two-sample t-tests were computed using the same non-parametric approach to evaluate between-group differences within this mask. Differences were considered significant with a p value < 0.05 after FDR correction. The mean z scores of the significantly altered DNFC between patients and controls were extracted for each subject and Pearson’s correlation was computed with the severity scale. Results were Bonferroni corrected at p < 0.05.