This dataset includes raw T1-weighted, diffusion MRI, and functional MRI (task and resting-state) scans from healthy controls and patients with focal dystonia (cervical dystonia and upper limb dystonia). Processed data may be available in a future release. All data were collected at Hospital Israelita Albert Einstein using a Siemens MAGNETOM Prisma 3.0 Tesla MRI scanner equipped with a 64-channel head coil. All data are de-identified and shared in compliance with ethical and legal guidelines. The dataset supports investigations into the neural correlates of focal dystonia. This dataset is made available under the Creative Commons Zero (CC0) v1.0 License.

Participants were adults aged 18 to 78 years, literate in Portuguese, and right-handed. Patients were diagnosed with idiopathic focal dystonia according to current criteria and were recruited from movement disorder clinics and public advertisements. Healthy control participants were free of any neurological or psychiatric conditions. Exclusion criteria included any other neurological or psychiatric diagnosis, prior neurosurgery, moderate-to-severe head trauma, cancer, uncontrolled metabolic conditions, current or past use of neuroleptics, history of substance abuse, or contraindications for MRI scanning such as claustrophobia or the presence of metal implants. All participants provided written informed consent, and the study was approved by the Institutional Review Boards of Hospital Israelita Albert Einstein, Hospital do Servidor Público Estadual, and Universidade Federal de São Paulo.

Clinical assessments included the Burke-Fahn-Marsden Dystonia Rating Scale (BFMDRS) for all patient participants. Demographic data were collected, including age and sex. Clinical information regarding dystonia groups [upper limb dystonia (ULD), cervical dystonia (CD), and healthy controls (HC)], BFMDRS scores, age at acquisition, and sex are provided in participants.tsv.

T1w: Structural MRI data were acquired using a T1-weighted MPRAGE sequence with the following parameters: repetition time = 2500 ms, echo time = 3.47 ms, inversion time = 1100 ms, flip angle = 7°, matrix size = 256 × 256, 192 sagittal slices with no gap, and isotropic voxel dimensions of 1.0 × 0.5 × 0.5 mm. Acquisition time was 2 minutes and 34 seconds.

DWI: Diffusion-weighted images were collected with 60 diffusion directions at a b-value of 1000 s/mm² and 9 interleaved b0 images. The phase-encoding direction was anterior–posterior, with an additional reverse phase-encoding acquisition for distortion correction. Diffusion parameters included repetition time = 3151 ms, echo time = 84.2 ms, voxel size = 2 mm isotropic, and 69 axial slices covering the entire brain.

fMRI (resting-state): BOLD echo planar imaging (EPI) was used with the following parameters: TR = 900 ms, TE = 33 ms, flip angle = 52°, matrix size = 64 × 74, slice thickness = 2 mm, gap = 0.6 mm, total volumes = 192. Resting-state scans lasted 6 minutes and 11 seconds. Participants were instructed to stay awake, relaxed, with eyes open and fixed on a projected bright crosshair on a dark background.

fMRI (finger-tapping task): BOLD EPI was used for a finger opposition task with the following parameters: TR = 2000 ms, TE = 25 ms, flip angle = 67°, matrix size = 64 × 74, slice thickness = 2.5 mm, gap = 0.6 mm, total volumes = 108. Acquisition time was 12 minutes. The task consisted of 12 blocks (four blocks per condition: right hand, left hand, both hands), each lasting 30 seconds, followed by 30-second rest intervals. Participants were instructed to tap each finger to the thumb, from index to little finger, repeating the sequence as quickly and accurately as possible. Blocks were randomized using E-Prime® 2.0 software. Task onset, duration, and condition details are provided in task-finger_events.tsv in the func folder.

Of the 112 participants initially enrolled, four were excluded due to incidental MRI findings and 16 due to technical artifacts. The final dataset includes high-quality imaging data from 37 healthy controls, 25 patients with cervical dystonia, and 29 patients with upper limb dystonia. Anatomical images were de-identified using Pydeface 2.0.2.