Study On Abnormal Brain Activity In Stroke Patients Based On Resting State EEG

Stroke,which belongs to nervous disease,is known for its high incidence and high mortality.Brain ischemia or hemorrhage leads to local necrosis of brain tissue which would cause loss of nerve function.This dissertation studies and analyzes the difference of brain activity between stroke and normal subjects by spectrum,source localization and brain network construction,which is based on EEG?The experiment collects EEG data from 41 subjects,which include 18 normal subjects and 23 stroke patients whose lesions is located on the left hemisphere.All EEG signal were pre-processed which includes filtering and removal of artifacts before analysis.The analysis is carried out by the follow three aspects:First,EEG spectrum is calculated by wavelet transform.Through a series method of selecting features which means statistical t-test,multiple test correction,lasso regression and support vector machine(SVM),the dissertation selected a few feature with significant difference between stroke and normal in spectral analysis.During that analysis,we got the facts that the patients whose lesions is located on the left hemisphere has left-right asymmetry,which performs that the spectrum of left hemisphere is higher.Besides,there are significant differences in occipital lobe associated with visual function and central area associated with motion function.In the channel level,phase locked value(PLV)was used to construct the brain functional network with small-world characteristic and network parameters are calculated.By Comparing stroke patients and normal subjects through the methods for selecting feature which are described above,the feature which has significant difference were chosen.By using these feature,we can get a high-accuracy classifier that can be used to identify stroke patients.At last,this dissertation analyses the source location and brain functional network in source level.In the source localization analysis,source are divided into 77 areas according to the Brodmann area.Through statistical tests and multiple test corrections,areas with significant differences are obtained,which is mainly in the occipital lobe related to visual function and the central region related to motor function.The brain functional network is constructed by data from source location and multiple thresholds.In the analysis,one parameter are proposed to quantify the relationship between and inside the brodmann area.Through statistical t-test and multiple test correction,the features with significant differences are obtained,which mainly are relationship between the brodmann 17,18,19 related to visual function and the brodmann 29,30 responsible for memory.Besides,a small part of the feature is responsible for motion functions.The research about differences of stroke patients and normal subjects in this dissertation starts from three aspects above.It got a few classification model with high accuracy and features which can shows the significant difference between stroke patients and normal subjects.Parts of these feature are interpretable,which can strengthen understanding of changes in brain of stroke patients.It certainly has reference for the diagnosis and evaluation of clinical stroke patients.