A robust deep learning-driven framework for detecting Parkinson's disease using EEG.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Parkinson's disease (PD) is a neurodegenerative disorder affecting motor function.
• Early and accurate diagnosis of PD is crucial for patient well-being and treatment efficacy.

Purpose of the Study:

• To develop a deep learning-based approach for detecting Parkinson's disease using electroencephalogram (EEG) signals.
• To enhance PD detection accuracy through advanced signal processing and machine learning techniques.

Main Methods:

• Refinement of EEG data using a channel attention module.
• Generation of time-frequency maps via wavelet scattering transform.
• Augmentation of data using a Generative Adversarial Network (GAN).
• Classification using a CNN-Transformer model trained on augmented time-frequency maps.

Main Results:

• The proposed deep learning model achieved a high accuracy of 99.52% for Parkinson's disease detection.
• Data augmentation using GAN improved the similarity of time-frequency maps for both PD patients and healthy controls.

Conclusions:

• Deep learning models, particularly CNN-Transformer architectures, show significant promise for accurate PD detection from EEG signals.
• The integration of channel attention, wavelet scattering transform, and GAN-based data augmentation offers a robust framework for neurodegenerative disease diagnosis.