Cortico-basal oscillations index naturalistic movements during deep brain stimulation

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computational Neuroscience

Background:

• The basal ganglia and sensorimotor cortex are crucial for motor control, but their network dynamics in Parkinson's disease (PD) and during deep brain stimulation (DBS) are not fully understood.
• DBS effectively treats PD hypokinetic symptoms, and sensing neurostimulators can record brain activity, offering potential for improved motor decoding.

Purpose of the Study:

• To investigate the contributions of cortical and subcortical motor network nodes to naturalistic movement in PD patients receiving DBS.
• To evaluate the impact of DBS on cortico-basal circuit dynamics and movement-related neural signals.
• To develop and validate machine learning models for predicting naturalistic movement using neural recordings.

Main Methods:

• Recorded over 530 hours of cortical and subcortical neural signals from 15 PD patients during daily activities with subthalamic nucleus (STN) or globus pallidus interna (GPi) DBS.
• Utilized synchronized wrist-worn accelerometers to quantify forearm speed and movement states.
• Developed machine learning models using spectral features from neural data to predict movement and forearm speed.

Main Results:

• Cortical beta oscillations (13-30 Hz) effectively distinguished between mobile and stationary states.
• Subcortical high beta movement-related desynchronization (MRD) and gamma movement-related synchronization (MRS) correlated with movement kinematics.
• Combined cortico-basal models achieved high accuracy (AUC > 0.85) and correlation (r > 0.68) in predicting movement states and speed, outperforming models using single-site data.
• Higher DBS current amplitudes reduced subcortical beta MRD and gamma MRS, negatively impacting subcortical model performance but not cortical or combined models.

Conclusions:

• Cortico-basal motor network nodes encode complementary kinematic information crucial for accurate and stable decoding of naturalistic movement during DBS.
• Machine learning models integrating cortical and subcortical signals offer robust movement prediction, supporting the development of closed-loop, adaptive DBS (aDBS) systems.
• Findings provide insights into neural biomarkers for motor control and potential therapeutic targets for movement disorders.