Revealing the dynamic correlation between neural and muscular signals using time-dependent granger causality analysis

Functional correlation between oscillatory neural and muscular signals during tremor can be revealed by Fourier-transform-based coherence estimation. The coherence value in a defined frequency range reveals the interaction strength between the two signals. However, coherence estimation does not provide directional information, preventing the further dissection of the relationship between the two interacting signals. We have therefore investigated functional correlations between the subthalamic nucleus (STN) and muscle in Parkinsonian tremor using adaptive Granger autoregressive modelling. During intermittent resting tremor we analysed the inter-dependence of local field potentials (LFPs) recorded from the STN and surface electromyograms (EMGs) recorded from the contralateral forearm muscles. We compared the Granger causality with coherence estimation, an adaptive Granger causality based on autoregressive modelling with a running window was used to reveal the time-dependent causal influences between the LFP and EMG signals. Our results showed that during persistent tremor, there was a directional causality predominantly from EMGs to LFPs corresponding to the significant coherence between LFPs and EMGs at the tremor frequency; and over episodes of transient resting tremor, the inter-dependence between EMGs and LFPs was bi-directional and alternatively varied with time. We conclude that the functional correlation between the STN and muscle is dynamic, bi-directional, and dependent on the tremor status.