A Novel EEG-Based Deep Approach for Diagnosing Alzheimer's Disease Using Attention-Time-Aware LSTM

Clinical settings typically diagnose Alzheimer's disease (AD) as the most common form of dementia. The electroencephalogram (EEG) has proven to be the most reliable method of detecting AD. The goal of this study is to use EEG data to identify AD and improve a computer-assisted diagnostic model. Initially, windowing is used to analyze the non-stationary EEG data. This study presents the attention-time-aware long-short-term memory (ATLSTM) model, which incorporates windowed EEG segments to generate deep features for categorization and extraction. Consequently, a decision-maker model has been constructed to classify three different types of AD based on the EEG signals using a novel deep structure. To determine the effectiveness of the proposed method, the F-measure, specificity, sensitivity, and total detection accuracy rate were measured. The results showed that the ATLSTM deep learning models are the most accurate at predicting AD, with an accuracy rate of 95.23%. Implementing the proposed method has demonstrated its importance as an additional tool for identifying potential biomarkers to assist in AD diagnosis. To verify and validate the developed decision-making model, Figshare data sets covering AD, mild cognitive impairment (MCI), and control (CO) are used. Furthermore, the suggested approach enables identification of both mild cognitive impairment and Alzheimer's disease, unlike existing comparable systems that use two-class detection.