State-Space Modeling of Cortical Beta Waves in Parkinson's Disease

We identified state space models to describe beta oscillations in motor cortex recordings in five Parkinson's Disease patients with and without concurrent cortical electrical stimulation. Previous work used autoregressive models to remove brain stimulation artifacts and detect instantaneous phase and frequency. This work establishes a linear time invariant model that offers improved accuracy in predicting beta oscillation power. Using this approach, the effect of brain stimulation can be modelled, further improving the prediction accuracy. Clinical Relevance-Accurate models can be used to develop closed-loop neuromodulation approaches that have the potential to improve treatment of Parkinson's and other diseases. This model can also be used to analyze the network connectivity among hidden states, which may help quantitatively study pathological changes.