The Epilepsy Mechanism Based On EEG Brain Network

Epilepsy is a neural system disease,which is characterized by repeatability and unpredictability.Because of its unpredictability,epileptic seizure may threaten the personal safety of the patients or others.Moreover,the repeated seizure of epilepsy will lead to the severe damage to the neural system of patients,as well as their daily behaviors and brain cognition functions.Therefore,the investigation of the potential mechanism that accounts for the epileptic seizure is now urgent.In our present study,we concentrated on locating epileptogenic zones and probing the dynamic network characteristics preceding clinical seizures in epileptic patients,which meant to provide the new insights into understanding the mechanism of epilepsy and further to find the effective treatment of epilepsy.The main works of this dissertation are as follows:1.The research on locating epileptogenic zones with interictal discharging EEG.In this study,the time-varying brain networks were constructed by adaptive directed transfer function to reveal the dynamic patterns during the interictal discharging.Moreover,an algorithm based on graph theory was proposed to detect the epileptogenic zone.The findings revealed that although the epileptogenic zones varied across epileptic patients,the algorithm based on time-varying networks consistently predicted the epileptogenic zones.Specifically,the epileptogenic zone was firstly activated prior to the discharge onset,and then it worked as a source to propagate the activity to other brain regions.In other words,the area that activated firstly prior to the interictal discharging might be the epileptogenic zone.And the epileptogenic zones detected by our method were consist with those determined by usual clinical means.2.The research on probing the dynamic network characteristics before clinical seizures.The EEG with time length of 1 hour before clinical epileptic seizure was used to explore the underlying mechanism of epilepsy.Firstly,the brain networks of all frequency bands at different time periods were constructed by coherence;then four clustering approaches were adopted to find the distinct clusters of networks.Here,to ensure the reliability of results,this process was repeated 1000 times for each clustering approach.Thereafter,the adaptive median filtering was used to uncover the dynamic networks before clinical seizures of epilepsy,and we finally obtained the discrimination ratio of preictal from interictal periods,where only those clustering with continuous duration of network states are regarded as the accepted clustering result.The findings revealed that the k-medoids clustering on β band performed the best compared to the other three approaches on all frequency bands concerned.Then,based on the results of kmedoids cluster on β,we calculated the network properties of two epilepsy periods,which shows that the significant smaller characteristic path length,as well as the larger clustering coefficient,global efficiency,and local efficiency corresponding to the preictal period could be found when compared to those of interictal period.In other words,the synchronism among brain regions during preictal period is enhanced,even though no obvious discharging could be found in EEG.And,the synchronization revealed by networks may index the presents of epilepsy seizure.In conclusion,in our present study,the epileptogenic zone and the dynamic network characteristics before clinical seizures were systemically investigated in the interictal and preictal periods by using brain network analysis on EEG.The findings of present study might be a useful tool to investigate the epilepsy,i.e.,locating epileptogenic zone and also provide the new insight into understanding the epilepsy seizures.