Seizure Onset Zone Localization For Focal Epilepsy From EEG

Epilepsy is the second most common neurological disorder in China,with unpredictable epileptic seizures causing spasm and loss of consciousness.Seizures are controllable with medi-cation in about 70%of cases.Surgical resection and nerve stimulation may help for the remaining patients with drug-resistant epilepsy.It is necessary to localize the seizure onset zone before surgery.The cause of most cases of epilepsy remains unknown,which makes diagnosis and treatment even more difficult.With the developments in electronics and computer science,it is of growing promise to localize the epileptic foci by signal processing technology.The non-invasive scale EEG(sEEG)with low spatial resolution can cover the whole skull,which makes it suitable to localize the seizure onset zone(SOZ)roughly.The invasive EEG(iEEG,ECoG or stereoEEG)with high spatial and temporal resolution usually covers only a small area,which makes it suitable to localize the seizure onset zone(SOZ)more accurately after a rough SOZ localization.This thesis focuses on SOZ localization for focal epilepsy,based on dynamic source imaging using sEEG or Riemann Manifolds using iEEG.(1)Standard electrode positions are used.First,separate signals into independent components(ICs)in the sensor space by idependent component analysis(ICA).Then,select the seizure compo-nents based on the difference in time-frequency representation between seizure component and non-seizure component.A 3-layer 3D distributed source model based on boundary ele-ment method(BEM)was used to model the brain source distribution.By solving the inverse problem of the source imaging,we can get the source signal and localization,namely the SOZ.We add the smooth constraint according to the characteristics of brain structure and physiological mechanism.(2)Ictal recordings using intracranial electrodes(e.g.,subdural strips,grids,or depth electrode arrays)are necessary after the SOZ localization using sEEG.We analysis the regional epilep-tic network dynamic by the functional connectivity of the iEEG.A new measure named log-density ratio(ρ1)introduced by us reveals that there is a key state in the latter half of the seizure period,which differents from the former half or the pre-seizure period.We cluster the seizure period into 2 states by kmeans algorithm based on Riemann distance.Then we localize the SOZ by the mean matrix of connectivity matrices in the second state.