Model-based Predictive Control Of Epilepsy State

As a kind of common neural disorder disease, epilepsy is caused by local brain lesions and harm to human health. At present, research on model is the most popular research direction to a lot of scientists who focus on epileptic disease control research.The target of these theory researches based on model is to find a control law which makes the specific performance indicators optimal, the most attractive method is the optimal control algorithm.The goal of this article is to predictive control epilepsy disease model, predictive control not only retains the characteristics of the optimal control with optimal performance index, but also has the characteristic of the online rolling optimization, and predictive control is a kind of closed-loop control which can improve the effect of the clinical diseases of the nervous system of the open loop control.The research of this paper is based on the calculation model, this is the fastest, the most simple and the most effective step to understand disease.In this article, we choose three models, the multidimensional conductance compartment model can represent epilepsy, a one-dimensional phase model and neural network model, we implement predictive control to control discharge of neurons and neural network model, specific studies include the following contents:First of all, in this thesis, the control of multi-dimensional conductance model PR model is achieved by using a two layer control algorithm- generalized predictive control combined with input-output linearization. Two kinds of control strategies have been implemented to PR model: the output term is soma membrane potential, but the input term is different, one is soma and the another is dendrite. Under both control strategies, the stable and desired control effect of PR model is achieved. At the same time, the control performance index of predictive control algorithm and which of two lonely with input-output linearization control algorithms are compared.Secondly, predictive control is applied to realize the phase control of a one-dimensional reduced phase model. Make use of the phase response curve of PR model to achieve phase model of normal and epilepsy state respectively. Realizing the phase control of epilepsy phase model with generalized predictive control algorithm combined with input-output linearization, makes the phase tracking the phase of normal state model. Achieving the control of neural system disease using a one-dimensional simplified model predictiveFinally, realizing the control of epilepsy state with predictive control. Researching the relationship between Epileptic disease and the small world network, the average field potential, Hindmarsh-Rose(HR) neuron small world network model is established. Control the average field potential of epilepsy small world network synchronous firing pattern to average field potential of normal state small world network asynchronous firing pattern. In order to demonstrate the effectiveness of the predictive control algorithm, we also control the asynchronous firing pattern to synchronous firing pattern, realizing the transformation of field potential of neural network.This thesis provides the theory basis of the control problem of neural system disease with hardware implementation, a train of thought for the treatment of neural system disease and important theoretical value for the in vitro, in vivo studies and clinical researches on neural system disease.