Research On The Iterative Control Of Neurons And Neural Network Dynamics

Neurons response to various external stimuli by generating different action potential coding, and thus affects the nervous system function and regulation. The study of encoding mechanism in nervous system and how to control its encoding features has become one of the research hotspots of neurosciences. At present, in order to obtain neuronal properties excellently, the closed-loop electrophysiology has become an important method for the encoding mechanisms study of neurons and nervous system. The abnormal firing patterns in neurons and neural ensembles are the main pathological mechanism of the major mental illness. Deep brain stimulation is an effective method for treatment of neurological diseases, such as Parkinson’s Disease, epilepsy. However, the various control methods are mostly open-loop stimulation at present. The open-loop stimulation has not a generality character, and the control effect is heavily dependent on the model parameters. Additionally, neural system is a nonlinear system, and accurate parameters of neural system are unavailable. Therefore, in this thesis, iterative learning control method is proposed to realize the closed-loop control of neurons and nervous system. The research content of this thesis mainly includes the following three aspects.First, in this thesis, the closed-loop control of neuron’s firing pattern is achieved using the iterative learning control for the first time. Using two different neuron models belong to two different excitabilities:Hodgkin-Huxley (HH) model that belongs to class I neural excitability and Morris-Lecar (ML) model that belongs to class II neural excitability, iterative learning closed-loop control of firing pattern is realized. Under the external disturbance, simulation analysis is conducted. Simulation results show that under different neuron excitabilities and different desired firing patterns, iterative learning control can effectively realize the closed-loop control of its firing pattern.Secondly, the closed-loop control of neural network’s field potentials is achieved using iterative learning control for the first time. Two kinds of neural network model are established:discrete mapping-based neural network model and the continuous Hindmarsh-Rose (HR) neural network model. For discrete mapping-based neural network model, iterative learning control is applied to realize mean-field potential conversion control between synchronization status and desynchronized status. For continuous HR neuron network model, tracking control of mean-field potential for different rewiring probability on the desired sine wave is realized.Finally, according to the pathogenesis and the clinical features of Parkinson’s disease, a simplified cortico-basal ganglia-thalamocortical model is established. Due to various research difficulties, such as the complexity of neural system, the effect of noises and the uncertainty of accurate parameters in TC relay neuron model, the iterative learning control (ILC) is proposed to realize the closed-loop control of parkinsonian state in thalamocortical relay neuron. Simulation results show that iterative learning control method can effectively restore hypothalamic neurons of Parkinson state relay reliability, thus effectively alleviate or cure motor symptoms of Parkinson’s disease.This thesis provides a train of ideas for the closed-loop electrophysiological studies of neurons or neural network. In addition, the control algorithm in this thesis can be directly applied to the closed-loop electrophysiological experiments or deep brain stimulation device.