Using Deep Pyramid CNN for Early Detection of Alzheimer’s Disease

This study proposes a deep learning framework for early Alzheimer’s disease (AD) detection using EEG signals, leveraging a novel feature extraction method, Empirical Fourier Decomposition (EFD). EEG data from ۱۰۰ subjects, including AD, mild cognitive impairment (MCI), and healthy controls (HC), were analyzed. Frequency-domain features extracted via EFD were compared with those from the Empirical Wavelet Transform (EWT). A ۱۶-coefficient feature vector, based on prior research, was used, and data were augmented through an overlapping sliding window technique. Classification with a Deep Pyramid Convolutional Neural Network (DPCNN) achieved ۹۸.۲۹% accuracy, ۹۸.۲۸% sensitivity, and ۹۹.۱۴% specificity. Results demonstrate EFD’s effectiveness over EWT in capturing discriminative EEG patterns, highlighting its promise for early and accurate AD diagnosis.

کلیدواژه ها:

Empirical Wavelet Transform ، Empirical Fourier Decomposition ، Deep Pyramid Convolutional Neural Network ، Alzheimer’s Disease