An efficient Parkinson's disease detection framework: Leveraging time-frequency representation and AlexNet...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Biomedical Engineering

Background:

• Parkinson's disease (PD) affects millions globally, necessitating early diagnosis for effective management.
• Electroencephalogram (EEG) signals offer potential for PD monitoring, but traditional methods lack regional specificity and real-time performance.
• Existing EEG-based PD detection methods require improvement in accuracy and efficiency.

Purpose of the Study:

• To develop a novel approach for efficient and accurate early diagnosis of Parkinson's disease using EEG data.
• To identify critical brain regions and electrode locations (e.g., AF4, AFz) that provide the most representative features for PD detection.
• To enhance the performance of EEG-based PD diagnosis for real-time applications.

Main Methods:

• Utilized a Time-Frequency Representation (TFR) combined with an AlexNet Convolutional Neural Network (CNN) model.
• Employed Wavelet Scattering Transform (WST) to capture temporal and spectral EEG signal characteristics.
• Conducted channel-based analysis to pinpoint significant brain regions for PD identification.

Main Results:

• The proposed AlexNet CNN model achieved high accuracy: 99.84% on the San Diego dataset and 95.79% on the Iowa dataset.
• Identified frontal and central brain regions, specifically AF4 and AFz electrodes, as crucial for PD detection.
• Demonstrated superior performance compared to existing EEG-based PD detection methods.

Conclusions:

• The novel TFR-based AlexNet CNN approach offers a significant advancement in early Parkinson's disease diagnosis.
• Frontal and central brain regions contain key EEG features vital for accurate PD identification.
• This research paves the way for essential technology to improve PD diagnosis, patient care, and quality of life.