Machine Learning Framework of EEG for binary classification of Early on-set Alzheimer’s Disease

Dementia is the one type of brain degenerative disease. It affects the memory, thinking and ability to perform daily task of a person who has dementia. Out of all dementia cases at least two-thirds of all dementia cases in people are above the age of 65. The most common cause of dementia is Alzheimer’s disease(AD) and this tells the importance of AD in the case of Dementia. In this paper Electroencephalogram (EEG) signal Dataset, is used to classify AD from Normal control(NC). Early-onset Alzheimer’s disease (EOAD) which occurs during age less than 65 is given less importance to study than the lateonset form (LOAD). But, EOAD mostly show, disease progression more aggressively than LOAD. So, this paper discusses about the binary classification of AD in EOAD age group whose age is less than 65 and LOAD age group whose is age greater than or equal to 65 and compare the classification accuracy of both. For binary classification, Support Vector Machine(SVM) is used which is a machine learning technique. Maximum alpha frequency band(812 Hz) power between Different Electrode Location of brain is used as a feature input to Support Vector Machine. Along with EEG dataset Mini Mental state Examination (MMSE) score of patients is given. Based on MMSE score, the label is generated and given as an input to SVM classifier. To ensure robust model performance K-fold stratified cross validation is used in SVM based binary classification. To measure the performance of the model, average folded accuracy is used. After studying the binary classification Accuracy results of EOAD group, it is founded that electrode location O1 and O2 which is in occipital brain lobe has more average folded accuracy which is 92.86% whereas LOAD age group has less average folded accuracy of 73.33% for same electrode Locations. This is evidently shows that maximum alpha frequency band between O1 and O2 electrode Location is act as a best feature to distinguish Alzheimer’s from Normal Control in EOAD age group than LOAD group.