Channel Stacking: A Rapid Classification Method for Parkinson's Disease Based on EEG Data.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Parkinson's disease diagnosis relies on clinical symptoms, often diagnosed late.
• Electroencephalogram (EEG) signals offer a non-invasive method for neurological disorder detection.
• Accurate and early detection of Parkinson's disease is crucial for effective management.

Purpose of the Study:

• To introduce and evaluate a novel channel stacking technique for Parkinson's disease identification using EEG signals.
• To develop a deep learning model capable of efficiently processing multi-channel EEG data for accurate classification.
• To assess the model's performance using rigorous cross-validation on clinical datasets.

Main Methods:

• A channel stacking technique was developed to create comprehensive input representations from multi-channel EEG signals.
• A ResNet18 deep learning network was employed for the classification task.
• Leave-One-Subject-Out Cross-Validation was utilized to validate the model's generalization performance.

Main Results:

• The proposed channel stacking method combined with ResNet18 achieved a high accuracy of 96.43% in identifying Parkinson's disease.
• The model demonstrated effectiveness using only 90 seconds of EEG data per subject.
• Leave-One-Subject-Out Cross-Validation confirmed the model's robust performance on clinical data.

Conclusions:

• Channel stacking is a promising technique for enhancing EEG signal representation in deep learning models.
• The developed deep learning approach offers a highly accurate and efficient method for early Parkinson's disease detection.
• The study highlights the potential of this method for real-world clinical application in diagnosing Parkinson's disease.