Computational Modeling of the Motor Cortex

Abstract

Brain Computer Interfaces (BCI’s) can greatly improve the lives of people with brain disorders and disabilities, such as Parkinson’s disease or locked-in syndrome. Computational models of the brain can improve the efficiency of these systems. In this thesis we attempt to build a functioning computational model of 6000 Izhikevich neurons representing the basal ganglia, thalamus and motor cortex. Compared to physiological data, the model is accurate in terms of average firing rates and spike distribution, and plausible in terms of frequency distribution. However, during a simulated task the model did not respond in accordance with the data. Additionally, this thesis explores the merits and pitfalls of automated optimization of model parameters. We found that a fitness function based on firing rates can lead to multiple global maxima, and suggest expanding the fitness function to include frequency and synchrony.