Timur Latypov, Alexandre Berger, Manuela Mohareb, Karim Mithani, Sara Breitbart, Lindsey M. Vogt, Thiemo Florin Dinger, Inge A. Meijer, Alexander G. Weil, A. Fasano, Carolina Gorodetsky, George M. Ibrahim
BACKGROUND: Dystonia in children is a heterogeneous condition with variable response to deep brain stimulation (DBS). Brain-age gap, a machine learning-derived metric of structural deviation from norm, may capture signatures that differentiate underlying biotypes and predict outcomes. METHODS: A brain age model was trained on several thousand normative developmental trajectories (n = 2623). Brain-age gap (the difference between predicted and chronological age) was computed from pre-DBS T1-weighted magnetic resonance imaging in 37 children with dystonia and compared to matched healthy controls. Associations with etiology, Burke-Fahn-Marsden Dystonia Rating Scale (BFMDRS) and Pediatric Quality-of-Life (PedsQL) scores were examined. RESULTS: Children with dystonia showed a greater brain-age gap (structural deviation) compared to controls (P < 0.001). Greater gap was linked to worse baseline motor scores and poorer 1-year quality-of-life improvement. Patterns differed by etiology, with distinct regional deviations in genetic and acquired dystonia. CONCLUSION: Brain-age modeling reveals biologically distinct subtypes of pediatric dystonia and may offer a biomarker for stratification and outcome prediction. © 2026 The Author(s). Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.