Monitoring Progression In Alzheimer's Disease With EEG Source Activity

Alzheimer's Disease(AD)is a neurodegenerative disease that places an incalculable mental and economic burden on families and society with the seriousness of aged population problems.Although intervention can be carried out at an early stage of the disease.However,there is a lack of economic and effective clinical detection methods.In recent years,Electroencephalogram(EEG)technology has been developed as an important means to detect abnormal brain activity in patients with Alzheimer's disease.In particular,the development of EEG reconstructed technology not only makes use of the advantages of high time resolution of EEG,but also basically solves the influence of volume conduction effect,which can be traced back to the activity of cerebral cortex.The special data set in this paper is composed of members of familial Alzheimer's disease(PSEN1 E280 A mutation),including three groups of subjects with different states.Among them,healthy non-carrier group(Ctrl),asymptomatic carrier group(ACr)and symptomatic carrier group(SCr),in which asymptomatic carriers have almost 100% probability of showing AD symptoms in the future.Familial Alzheimer's Disease(FAD)provides an excellent opportunity to better explore effective biomarkers of Alzheimer's disease.In order to better explore the effective biomarkers of Alzheimer's disease,this paper will study the biomarkers that affect the development of Alzheimer's disease from the following two aspects:(1)Applying the innovative EEG reconstructed method BC-VARETA(Brain Connectivity Variable Resolution Tomographic Analysis)model,which is accurate in location and good in robustness,compared with eLORETA,it can better solve the influence of ”leakage effect” caused by volume conduction when source activity and connectivity are estimated.The BC-VARETA model was used to analyze the inverse problem of EEG data of three groups of patients with familial Alzheimer's disease in different stages,and the more accurate source activity of EEG stationary time series in frequency domain was calculated.It provides a good basis for the screening of biomarkers.(2)The activity feature matrix of brain power supply based on BC-VARETA traceability model is classified by using the Elastic-Net stabilization classifier based on prescreening proposed in this paper.After enough iterations,the selected significant biomarker features and the AUC values classified by these features are obtained.The classification results show that the entorhinal cortex at 9.39 Hz can be used as a biomarker to distinguish the three groups,and the AUC value of the three classification is 0.93,and a good classification effect is obtained.At the same time,the fusiform cortex had a better classification effect in distinguishing from in asymptomatic carrier(ACr)group and symptomatic carrier(SCr)group.It is possible to provide theoretical basis and methods for the diagnosis and prediction of Alzheimer's disease by analyzing the differences of the most screened biomarkers in the three groups of different subjects.To sum up,on the one hand,this paper applies the innovative EEG reconstruction algorithm to extract the activity features of EEG effectively,which provides a new idea for the analysis of EEG inverse problems.On the other hand,it optimizes the classification effect of Elastic-Net classification algorithm for high dimensionality,small sample and multi-packet data,improves the stability and classification effect of the classification algorithm,and finally finds the ideal biomarkers.Through the screening of biomarkers by reconstruction algorithm and classification algorithm with stability and robustness,the characteristics that affect Alzheimer's disease at different stages were discovered.It provides theoretical basis and practical direction for the prediction and diagnosis of Alzheimer's disease in the future.