Study On Parameter Estimation And Simulated Regulation In A Neural Mass Model Of Epilepsy

Epilepsy,a common brain disease with transient central nervous system dysfunction,is characterized by recurrent seizures that are hard to be predicted.Epileptic seizures not only seriously affect the normal life of patients,but also put heavy burden on the society.Therefore,effective modulation of epileptic seizures has becoming more and more important in neuroscience.Accumulating evidence suggests that epileptic seizures are mainly resulted from imbalance between neural excitation and inhibition in the central nervous system.Accordingly,if the excitability and inhibitory values can be measured in real time,it will provide the possibility of effective closed-loop control of seizures.Thus,in this thesis,we will combine electroencephalograph(EEG)signals with a neural mass model to estimate several model gain parameters,with which designing a closed-loop controller to detect and modulate epileptic seizures.Overall,three main findings of this thesis are shown in the following:(1)Combining the Unscented Kalman Filter(UKF)with the Jansen & Rit neural mass model to estimate model parameters.First,we simulated various neural oscillations with different model parameters,and found that the model could mimic epileptiform discharges with changing the excitatory and inhibitory synaptic gains.These findings suggest that it is reasonable to employ the neural mass model to uncover an approach of controlling epileptic seizures.Then,by comparing the advantages and disadvantages of several commonly used Kalman filters,it is determined that the Unscented Kalman Filter is used for EEG signal filtering and model parameter estimation.(2)Estimating the excitatory and inhibitory synaptic gain parameters with UKF.Based on the UKF method,we successfully estimated the excitatory and inhibitory synaptic gain parameters with the simulated EEG signals.Furthermore,we extended this method to the real EEG signals obtained from Children's Hospital Boston and Massachusetts Institute of Technology.We found that the ictal and interictal periods could be clearly distinguished with those estimated model gain parameters,which might provide a new way to diagnose epileptic seizures.(3)Designing a closed-loop control method to suppress epileptic seizures.First,we used the traditional proportional-integral(PI)closed-loop control method to modulate epileptic seizures.Although we could completely terminate epileptic seizures,selecting optimal PI control parameters is tedious and time-consuming.Therefore,in order to simplify the parameter optimization process of PI closed-loop control method,we designed an improved PI closed-loop control method,which has been confirmed better than the traditional PI closed-loop control method on simulated EEG signals.Especially,the rigid selection of optional PI control parameters has been discarded.In summary,the improved PI closed-loop control method proposed in this thesis not only provides new sights into the closed-loop control of epileptic seizures in the future,but also offers theoretical guidance for clinical treatment of epilepsy.