Characteristics Analysis And Closed-loop Control Of Neuronal Networks

Neuroscience is a discipline that explores the causal relationship between neuronal activity and its corresponding behavior,one of whose cores is to obtain the desired output characteristics by stimulating neurons and networks.However,due to the high nonlinearity of the neural system,the same stimulus may generate different outputs,and different stimulus may generate the same output.Therefore,it is very difficult to obtain the desired output characteristics of neurons and networks.Although the advance in neural recording methods makes it possible to obtain a large amount of neural data,the precise ‘stimulus-response' relationship of neurons and networks cannot be derived without knowing anatomical structure.In addition,anomalous variation in neuron and network output characteristics can result in dysfunction of the neural system,leading to the occurrence of various mental illnesses such as epilepsy and Parkinson disease(PD).As one of the most important therapeutic methods for mental illnesses,electromagnetic stimulation can not adjust stimulating parameters according to patient's states as well as individual variation which causes side-effects and overconsumption of energy.Taking computational models of single neuron,neural network and neural population as a bridge connecting neural data and anatomical structure,this thesis constructs closed-loop systems which extract biomarkers depicting the output behavior and adjust stimulating parameters with different control algorithms to obtain expected outputs.The content of this thesis mainly includes:(1)Control of single neuron output characteristics: Single neuron is the basic unit of the neural system,and its output characteristics affect network characteristics as well as the function of the neural system.Based on the idea of voltage and dynamic clamp,the concept of ISI clamp is proposed which adopts ISI to depict different output characteristics,such as firing frequency and firing patterns,and constructs a closed-loop system to obtain expected output characteristics.The unscented Kalman filter is utilized to estimate ion channel characteristics of sodium,potassium and chlorin.In this way,neural activities are reconstruction,influences of changes in ion channels on firing activities are obtained and the pathogenesis of epilepsy and PD in single neuron level are simulated.(2)Control of the neuronal network output characteristics: The minimum CPG network is a basic functional unit of the neuronal network to realize different rhythms such as sleep,respiration,movement and digestion.Controlling its output characteristics is helpful to understand the mechanism and functional characteristics of network rhythms.In this thesis,the classic single neurons with morphological characteristics are adopted to construct the minimum CPG network,and firing regularities are selected to depict output characteristics.The closed-loop system is constructed to realize the control of irregular firing according to influences of neural parameters and connecting characteristics on output,and to get the desired output characteristics of the network as well as the relationship between network parameters and rhythm for simulation of rhythms associated with epilepsy and PD.(3)Control of abnormality neuronal network characteristics-epileptiform firing: The individualized therapy can be realized by adjusting the parameters of electromagnetic stimulation according to the changes of clinical conditions and the characteristics of electrophysiological signals.Neural mass model produces abnormal rhythm during epileptic state.The characteristics of the oscillation are extracted to depict the epileptic states,and an adaptive fuzzy closed-loop system is constructed to realize real-time and personalized control of epileptiform firing.This closed-loop system provides a feasible solution to the problem that precise models of individual differences and disease states cannot be obtained in physiological experiments.Simulation results show that the closed-loop control is effective.(4)Control of abnormality neuronal network characteristics-PD: The selection of stimulating targets seriously affects therapeutic effect of electromagnetic stimulation.Based on the pathogenic mechanism and clinical features of PD,the neural mass model was used to construct a computational model of PD focal area.A closed-loop system with the external globus pallidus as a stimulus target and its local field potentials as feedback variables was designed to control PD rhythm.This research indicates that the external globus pallidus could be a new target for electromagnetic stimulation.Results of this thesis provide new ideas for exploring basic electrophysiological problems,developing new disease symptoms and rehabilitation therapy.