Preoperative Functional Imaging Predicts Deep Brain Stimulation Outcome in Craniocervical Dystonia.

Abstract

Background: Deep brain stimulation (DBS) is an established treatment for craniocervical dystonia (CCD), yet postoperative benefit varies widely and cannot be estimated before surgery. Resting-state functional MRI (rs-fMRI) has shown prognostic value across DBS indications. The objective of this study was to develop and externally validate preoperative rs-fMRI models estimating DBS response in CCD.

Methods: Forty-two patients from two independent centers underwent preoperative 3 T rs-fMRI (discovery, n = 26; validation, n = 16). Using 12 literature-informed, symptom-relevant regions of interest, we quantified the amplitude of low-frequency fluctuation as a measure of intrinsic regional neural activity. We applied regularized modeling to classify super-responders, distinguish responders from non-responders, and estimate improvement in Burke-Fahn-Marsden Dystonia Rating Scale motor subscore.

Results: When applied to the validation cohort, the imaging models achieved a three-category classification accuracy of 75.0% (50.5%-89.8%) using a sequential two-step procedure (P_perm = 0.002). Predicted continuous improvement was also correlated with empirical improvement (R_ext = 0.65, Q² = 0.34, P_perm = 0.004). Within the discovery cohort, nested leave-one-out-cross-validation, in which feature selection and decision-threshold derivation were repeated independently within each training fold, yielded areas under the curve (AUCs) of 0.73 (0.50-0.96) and 0.77 (0.51-1.00). Across models, baseline primary motor cortex activity was consistently associated with diminished postoperative benefit.

Conclusions: Preoperative rs-fMRI captures clinically relevant variation in baseline brain state in CCD, with predictive performance supported in an external validation. These findings provide preliminary support for rs-fMRI as a potential source of information to aid patient counseling and clinical decision-making, pending prospective validation. © 2026 International Parkinson and Movement Disorder Society.