eva‎luation of Functional and Effective Connectivity methods in the diagnosis of Mild Cognitive Impairment using EEG signal

10/14/2024 12:00:00 AM

Alzheimer's disease is a type of neurodegenerative disease that is one of the most common types of dementia. The rapid growth of this disease has increased the need of the human society to find a reliable method for the early diagnosis of patients in the early stages of Alzheimer's, which is called mild cognitive impairment. Being in different perceptual, cognitive and emotional situations is associated with a kind of information dissemination through brain neuron oscillations. Many studies have investigated brain connections in people with various diseases such as Alzheimer's, schizophrenia, etc., but fewer studies have been conducted on patients with mild cognitive impairment. In this study, we intend to distinguish people with mild cognitive impairment from healthy people based on the resting EEG data using the communication analysis of different brain regions based on the analysis of functional communication and affective communication. In this thesis, the decoding of the brain signals of 27 participants, including 11 people with mild cognitive impairment and 16 healthy people, obtained from the recording of 19-channel EEG signals while resting, has been used. The purpose of this study is to separate and classify two groups of healthy people and MCI patients in extracting features from electroencephalogram signals based on power features of frequency bands and variance and features based on functional and effective communication using two average and non-average approaches. is taking. To select the best features and reduce the dimension of the feature matrix from different feature reduction methods, including the mutual information method, sorting the features with the PCA method and adding them, sorting the features with the t_test method and adding them, and the PCA method and thresholding. 2 classification methods including LDA and SVM with three polynomial, RBF and gussian kernels were used for averaging mode and 3 classification methods including LDA, KNN and SVM with two polynomial and gussian kernels for non-averaging mode and Leave_one_subject_out eva‎luation method were used. The best performance in feature extraction methods was related to functional communication methods, which correlation as a measure of functional communication methods has the best performance with LDA classification and 95% accuracy rate.

الكتروانسفالوگرافي , اختلال شناختي خفيف

EEG , MCI

Maryam Sadat Ojagh Kazazi

Mohammad Reza Daliri