Neuronal Complexity in Subthalamic Nucleus is Reduced in Parkinson’s Disease

Several theories posit increased Subthalamic Nucleus (STN) activity is causal to Parkinsonism, yet in our previous study we showed that activity from 113 STN neurons from two epilepsy patients and 103 neurons from nine Parkinson's disease (PD) patients demonstrated no significant differences in frequencies or in the coefficients of variation of mean discharge frequencies per 1-s epochs. We continued our analysis using point process modeling to capture higher order temporal dynamics; in particular, bursting, beta-band oscillations, excitatory and inhibitory ensemble interactions, and neuronal complexity. We used this analysis as input to a logistic regression classifier and were able to differentiate between PD and epilepsy neurons with an accuracy of 92%. We also found neuronal complexity, i.e., the number of states in a neuron's point process model, and inhibitory ensemble dynamics, which can be interpreted as a reduction in complexity, to be the most important features with respect to classification accuracy. Even in a dataset with no significant differences in firing rate, we observed differences between PD and epilepsy for other single-neuron measures. Our results suggest PD comes with a reduction in neuronal “complexity,” which translates to a neuron's ability to encode information; the more complexity, the more information the neuron can encode. This is also consistent with studies correlating disease to loss of variability in neuronal activity, as the lower the complexity, the less variability.