Dynamic neural states underpin bradykinesia severity in Parkinsons disease

BackgroundBradykinesia in Parkinsons disease (PD) may arise due to transient, network-wide neural dynamics that extend beyond beta-band oscillatory activity within the motor cortical-subthalamic nucleus (STN) circuit. MethodsWe address this question by using Hidden Markov Models (HMMs) to identify neural states from chronic motor cortical and STN recordings in five PD patients (1,046 hours from 10 hemispheres), with concurrent measurements of bradykinesia using wearable sensors. FindingsWe identified four neural states with distinct spectral and temporal features. Two states exhibited spectral signatures--particularly STN low and high gamma, STN delta/alpha, cortical beta, and cortico-STN beta coherence--that predicted worsening bradykinesia. However, STN beta power alone was not consistently predictive, challenging traditional beta-centric views. These states also displayed compensatory features associated with bradykinesia amelioration, including cortical delta/alpha activity, cortical high gamma, and cortico-STN high gamma coherence. Two additional states affected bradykinesia through temporal rather than spectral properties. Prolonged lifetimes of one of these state worsened symptoms, whereas increased occurrences of another, marked by local beta without cortico-STN beta coherence, improved motor function. InterpretationOur findings highlight the multidimensional nature of bradykinesia and suggest that state-aware, adaptive interventions targeting state features--rather than single frequency bands--may offer new opportunities for improved deep brain stimulation in PD. FundingAO is supported by an MRC Clinician Scientist Fellowship (MR/W024810/1) and a Rosetrees Trust/Race Against Dementia Team award. BA and AO acknowledge funding support from the Oxford University Hospitals Charity and Jon Moulton Trust. TL acknowledges funding support from the China Scholarship Council. Research in ContextO_ST_ABSEvidence before this studyC_ST_ABSPrevious work has demonstrated that subthalamic nucleus oscillatory activity at beta (15-30 Hz) frequencies correlates positively with motor symptoms in Parkinsons disease. This has led to beta activity being used as a biomarker for adaptive Deep Brain Stimulation. It remains unclear however whether other oscillatory features within the broader motor cortical-subthalamic nucleus network could provide improved biomarkers for tracking symptom severity. Added value of this studyWe address this by performing chronic motor cortical and subthalamic nucleus recordings in Parkinsons disease patients during activities of daily living. Simultaneous measurements of symptom severity were captured using wearable sensors. We used Hidden Markov Models to identify transient states of neural activity and related these to symptom severity. Although cortical beta and cortico-STN beta coherence predicted worsening motor symptoms, STN beta activity alone was not a consistent predictor. Interestingly, we identified spectral features associated with motor symptom improvements, including cortical delta/alpha activity, cortical high gamma, and cortico-STN high gamma coherence. Additionally, there was a compensatory state characterised by short-lived cortical and subthalamic nucleus beta activity, whose increased occurrence was associated with symptomatic improvements. Implications of all the available evidenceOur findings highlight the importance of prolonged, high temporal resolution measurements of both neural activity and symptom severity for discovering adaptive Deep Brain Stimulation biomarkers. Crucially, we identify new target states and spectral features for improving motor symptoms in Parkinsons disease.