Sensitivity analysis-guided Bayesian Parameter Estimation for Neural Mass Models: Applications in Epilepsy

Abstract

AbstractIt is well established that neural mass models (NMMs) can effectively simulate the mesoscopic and macroscopic dynamics of electroencephalography (EEG), including epileptic EEG. However NMMs are characterized by a high-dimensional parameter space and a lack of knowledge on what NMM parameters can be reliably estimated, thus limiting their application to clinical EEG data. In this article, we analyze the parameter sensitivity of Jansen and Rit NMM (JR NMM) in order to identify the most sensitive NMM parameters for reliable parameter estimation from EEG data. We also propose a joint estimation method for NMM states and parameters based on expectation–maximization combined with unscented Kalman smoother (UKS-EM). Global sensitivity analysis methods including Morris method and Sobol method are used to perform sensitivity analysis. Results from both the Morris and Sobol method show that the average inhibitory synaptic gain,Band the time constant of the average inhibitory post-synaptic potentials (PSPs),b−1have significant impact on the JR NMM output along with having the least interaction with other model parameters. The UKS-EM method for estimating the parametersBandbis validated using simulations under varying levels of measurement noise. Finally we apply the UKS-EM algorithm to intracranial EEG data from 16 epileptic patients. Our results, both at individual and group-level show that the parametersBandbchange significantly between the pre-seizure and seizure period, and between the seizure and post-seizure period, with transition to seizure characterized by decrease in averageBand high frequency activity in seizure characterized by an increase inb. These results establish sensitivity analysis guided Bayesian parameter estimation as a powerful tool for reducing the parameter space of high dimensional NMMs enabling reliable and efficient estimation of the most sensitive NMM parameters, with the potential for online and fast tracking of NMM parameters in applications such as seizure tracking and control.