Machine Learning Based Classification for Alzheimer's Disease Using EEG Signal

Alzheimer’s disease (AD) is neurodegenerative disorder that affects memory, cognition, and behavior,
representing the most common cause of dementia with old people. Alzheimer’s disease detection was
performed using convolutional neural networks (CNNs) applied to electroencephalogram (EEG) signals,
where the features were extracted from signal transformation coefficients. The EEG dataset collected from
48 participants, divided into two groups: Alzheimer’s disease patients and healthy controls. Several signal
transformation techniques were compared, including the fast Fourier transform (FFT), short-time Fourier
transform (STFT), synchrosqueezed Fourier transform (SSFT), continuous wavelet transform (CWT),
discrete wavelet transform (DWT) for 1D and 2D, and synchrosqueezed wavelet transform (SSWT), to
determine the most effective approach for EEG-based classification. Experimental results demonstrated that
the STFT method provided the highest performance, achieving superior accuracy, precision, sensitivity,
specificity, and F1-score for Alzheimer’s disease detection.