Analysis And Data-driven Modeling Of Epileptic Activity

Epilepsy is a neurological disorder affecting 50 million people worldwide,which is a heavy burden on patients and society.Predicting and controlling epilepsy is one of the important research directions of neuroscience.Computational modeling is of great significance to understand,predict and suppress epileptiform activity.Computational modeling provides an efficient way of structuring detailed knowledge and multi-modal data coming from research in neurobiology and neurophysiology.Computational models closely related with either experimental or clinical data could markedly advance our understanding of how and why seizures start,spread and stop within a restricted or more extended part of the brain.As a consequence,a data-driven model inversion framework of brain function imaging is introduced in this manuscript to reproduce,predict and control epileptic seizure.(1)Creating neural mass model(NMM)from electroencephalograph(EEG).Special emphasis is done to analyse parameter changes before and during seizures.An unscented Kalman filter is utilized to estimate parameters and states of the NMM in real time from the observed EEG.The information provided by a model-based framework could also predict seizures,and be used to provide feedback for electrical stimulators for robust prevention of seizures.(2)We first analyse the EEG obtained from an experiment of childhood absence epilepsy in the time domain and frequency domain,and then use a computational modeling approach combined with the electrophysiological data to infer about changes that may lead to a seizure.The advantage of using neural states and parameters as features for seizure prediction is that they are naturally patient-specific.The reconstructed EEG signals exactly match the objective EEG,and the estimated parameters are closely connected with physiological states.The preferred model would be the one that is able to predict the hidden properties from the real EEG measurement signals.(3)Assessment of the effective connectivity among different brain regions during seizure is a crucial problem in neuroscience today.The model inversion framework is based on approximating brain networks using a multi-coupled NMM.Particle swarm optimization method is used to estimate the effective connectivity variation(the parameters of NMM)and the epileptiform dynamics(the states of NMM)that cannot be directly measured using electrophysiological measurement alone.The estimated effective connectivity includes both the local connectivity parameters within a single region NMM and the remote connectivity parameters between multi-coupled NMMs.As an additional value,the use of data-driven modeling methods will help to identify the epileptogenic zone.(4)This work proposes a closed-loop electrical stimulation strategy with key parameters feedback.This strategy includes the neural mass model,the patient-specific seizure prediction step,and the controller.When the epileptiform activities are estimated,the controller outputs control signal so that the epileptiform spikes can be inhibited immediately.Numerical simulations are carried out to illustrate the effectiveness of the proposed framework.Such biological data motivated online estimator and controller design will likely revolutionise how we can image the underlying activity of the epilepsy and devise improved methods for neurological monitoring,control and treatment.