The Study Of EEG-Based Stroke Assessment Methods And Lesion Localization

Stroke is already the third most deadly disease in China,and Incidence rate is increasing at a rate of 6% every year.In stroke patients,ischaemic stroke patients account for the majority.The electroencephalogram(EEG)reflects extracellular currents mainly resulting from excitatory and inhibitory postsynaptic and it is very sensitive for detecting cerebral ischaemic,hence various EEG abnormalities typically manifest following ischaemic stroke.Therefore,the study of EEG-based ischaemic stroke and the location of cerebral infarction are of great significance.Firstly,in the pre-processing process,the 50 Hz power frequency noise,motion artifact noise,and volume conduction effect interference are mainly filtered out.An adaptive filter is designed to filter out the power frequency in 50 Hz.For the motion artifacts,the eeglab-based FastICA algorithm and the empirical modal decomposition algorithm are used to compare the performance of the two algorithms.Through the performance of signal-to-noise ratio,correlation coefficient and mean square error,it is concluded that the eeglab-based FastICA algorithm is better for filtering motion artifacts.In this paper,the Lineariy Constrained Minimum Variance(LCMV)spatial filter is used to solve the volumetric conduction effect(VC).Secondly,this paper analyzes the EEG data of patients with ischaemic stroke by power spectrum,approximate entropy and sample entropy.Two sets of control experiments were set up,the first group of control experiments: patient group and healthy group;the second group of control experiments: patients before and after treatment.In the control experiment,it was found that the patient’s characteristic index can be evaluated.In the first group of control experiments: the delta(1-4Hz)band power value of the EEG in the location of the cerebral infarction lesion was higher than that of the normal person,and the power value of the alpha(8-13Hz)band was lower than that of the normal person.The cerebral infarction lesion location delta band sample entropy is lower than normal.In the second group of controlled trials: patients with improved symptoms after thrombolytic therapy,the delta band power spectrum value decreased,the alpha band power spectrum value increased,and the delta band sample entropy also increased.Conversely,for poor results,these changes are not obvious or unchanged.Through these characteristic information,it is possible to evaluate the effect of the patient’s thrombolytic therapy.Finally,the boundary element method is used to construct the patient head model.The location of the lesion was predicted according to the calculated characteristic indexes.On this basis,the location of the cerebral infarction lesion was estimated using the spatial filtering method(Beamformer)and the precise low-resolution electromagnetic tomography(eLoreta).The estimation results are compared with the final nuclear magnetic results,and the Beamformer method based on the dipole model can achieve lesion localization in the localization.