Movement decoding using spatio-spectral features of cortical and subcortical local field potentials

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Commercially available sensing-enabled deep brain stimulation (DBS) devices offer future potential for real-time, adaptive therapy.
• Machine learning (ML) models can leverage multi-channel brain recordings to individualize treatment.
• Spatial information from recordings may enhance ML model performance for therapy adaptation.

Purpose of the Study:

• To compare decoding performance using single-channel versus spatial filtering techniques.
• To investigate the feasibility of spatial filtering in invasive neurophysiology.
• To assess the utility of combined electrocorticography (ECoG) and subthalamic local field potentials (LFPs) for decoding grip force.

Main Methods:

• Intracerebral multitarget electrophysiology in Parkinson's disease patients undergoing DBS.
• Comparison of decoding performance from single channels versus spatial filtering.
• Utilized electrocorticography (ECoG) and subthalamic local field potentials (LFPs) for grip-force decoding.

Main Results:

• Spatial information improved decoding performance by 6%.
• The benefit of spatial filtering and the resulting spatial patterns were highly individual.
• Spatial filters provided meaningful neurophysiological insights into brain networks.

Conclusions:

• Spatial filtering can enhance brain signal decoding for deep brain stimulation.
• Individualized approaches are crucial for optimizing brain signal decoding.
• Multielectrode recordings and spatial filtering advance precision medicine for clinical brain-computer interfaces.