Abstract
We investigated the impact of visual states on basal ganglia oscillatory biomarkers, comparing local field potentials (LFPs) dynamics between Parkinson's disease (PD) and dystonia and developing a decoding model for state identification. Simultaneous LFPs recordings from the subthalamic nucleus (STN) or globus pallidus internus (GPi), and cortex were obtained from 18 PD and 18 dystonia patients. In the eyes-closed state, theta and alpha power increased in the basal ganglia, with stronger coherence to the central cortex, more pronounced in the STN than in the GPi. Machine learning models identified the eyes-closed state with 88% accuracy for STN and 77% for GPi. The sensorimotor STN and GPi were most informative. The present findings provide proof-of-concept that basal ganglia LFPs can reliably predict a physiological state, highlighting the potential influence of physiological oscillatory activity on pathological bands and its relevance for adaptive stimulation paradigms.