Research on Feature and Channel Selection of Alzheimer’s Disease EEG Signals Using the OBL - GWO Algorithm

Alzheimer’s disease (AD) is a common neurodegenerative disorder that causes damage to brain neuron connections and impairments in neural signal transmission. To explore new diagnostic methods for it, this study proposes an opposition-based learning gray wolf optimization algorithm (OBL-GWO). Firstly, the SODP method is used for feature extraction of EEG signals. Secondly, the OBL-GWO algorithm is applied for feature selection to obtain 8 features. SHAP analysis is conducted on the model and analysis of variance is performed on the features to verify the reliability of the selected features. Thirdly, the features are subjected to PCA for dimensionality reduction and fusion, and feature vectors are used to represent different EEG channels. Finally, the OBL-GWO algorithm is used for channel selection, and 11 optimal channels are selected, with a classification accuracy rate of 87.43%. After visualizing the channel distribution, it is found that the frontal lobe area shows significant differences between AD patients and normal individuals. This study provides a classification framework with high accuracy and low complexity for the early diagnosis of AD.