Classification of Dementia EEG Signals by Using Time-Frequency Images for Deep Learning

• Published in: Innovations in Intelligent Systems and Applications Conference (Online) pp. 1 - 6
• Main Authors: Sen, Sena Yagmur, Cura, Ozlem Karabiber, Akan, Aydin
• Format: Conference Proceeding
• Language: English
• Published: IEEE 11.10.2023
• Subjects: Alzehimer's Dementia (AD); Brain modeling; Classification; CNNs; Deep learning; Electroencephalography (EEG); Fourier transforms; Image segmentation; Intrinsic Time Scale Decomposition (ITD); Neurological diseases; Short-Time Fourier Transform (STFT); Spectrogram; Time-frequency analysis; Training data
• ISSN: 2770-7946
• DOI: 10.1109/ASYU58738.2023.10296777

Dementia is a prevalent neurological disorder that impairs cognitive functions and significantly diminishes the quality of life. In this research, a deep learning method is introduced for detecting and monitoring Alzheimer's Dementia (AD) by analyzing Electroencephalography (EEG) signals. To accomplish this, a signal decomposition technique known as Intrinsic Time Scale Decomposition (ITD) is employed to classify EEG segments obtained from both AD patients and control subjects. The analysis specifically concentrates on 5-second EEG segments, utilizing ITD to extract Proper Rotation Components (PRCs) from these segments. The PRCs are subsequently transformed into Time-Frequency (TF) images using the Short-Time Fourier Transform (STFT) spectrogram. These TF images serve as training data for a 2-Dimensional Convolutional Neural Network (2D CNN). The proposed approach is compared with the classification of the spectrogram of 5-second EEG segments using the same CNN architecture. The experimental results conclusively demonstrate the superior classification performance of the ITD-based approach when compared to the utilization of raw EEG signals.