Macroscale Gradient‐Informed Neural Oscillation Topography in Parkinson's Disease

Background Parkinson's disease (PD) is characterized by large‐scale disruptions in beta and gamma oscillations. Although subcortical beta power is an established biomarker for current adaptive deep brain stimulation (aDBS), it may not fully capture the global pathophysiological burden and the macroscale hierarchical reorganization of the cortex. Objective We characterize the frequency‐specific reorganization of the cortical hierarchy across resting and motor states using functional gradients. We sought to identify topographic biomarkers that emerge across different behavioral states and determine whether these hierarchical features provide predictive power for global motor severity. Methods High‐density electroencephalography and magnetic resonance imaging–based source reconstruction were employed in patients with PD (n = 35) and healthy control subjects (n = 34). To characterize cortical connectivity transitions, we applied a manifold learning framework to derive frequency‐specific functional gradients. We quantified the diagnostic and predictive utility of these hierarchical features and performed transcriptomic enrichment analysis to validate the biological relevance of the alterations. Results Patients with PD exhibited a macroscale reorganization of the cortical hierarchy that was both frequency specific and state dependent. These gradient‐based biomarkers effectively differentiated patient groups and significantly predicted global Unified Parkinson's Disease Rating Scale Part III severity. Findings showed a robust framework with distinct topographical signatures, manifesting as a redistribution of informative signals across cortical regions. Conclusions This work demonstrates that PD induces a macroscale reorganization of the cortical hierarchy. State‐dependent topographical biomarkers effectively predict clinical severity and align with the disease pathological landscape. By identifying optimal sensing sites across distributed networks, our findings provide a principled reference to support next‐generation, cortical‐guided aDBS.