A Novel Fusion Framework Combining Graph Embedding Class-Based Convolutional Recurrent Attention Network with Brown...

Area of Science:

• Neurology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Parkinson's disease recognition (PDR) is crucial for timely intervention, but EEG signal noise and variability pose challenges.
• Traditional deep learning methods often fail to capture complex temporal and spatial EEG data dependencies, limiting PDR accuracy.
• Existing PDR models struggle with precision, necessitating advanced techniques for improved diagnostic capabilities.

Purpose of the Study:

• To introduce a novel fusion framework, GECR2ANet+BBOA, for highly accurate EEG-based Parkinson's disease recognition.
• To overcome the limitations of traditional methods in analyzing noisy and non-stationary EEG signals for PDR.
• To enhance early detection of Parkinson's disease and improve overall diagnostic efficiency.

Main Methods:

• EEG signal preprocessing using WGIF-EEW for noise removal and entropy weighting.
• Feature extraction via IVGG19-GTAN to capture intricate spatial and temporal EEG dependencies.
• Classification using GECR2ANet optimized by BBOA for enhanced Parkinson's disease recognition accuracy.

Main Results:

• The GECR2ANet+BBOA model achieved superior performance, reaching 99.9% accuracy on the UNM dataset and 99.8% on the UC San Diego dataset.
• Achieved high sensitivity (99.4% and 99.1%) and F1-scores (99.3% and 99.2%) on both datasets, significantly outperforming prior methods.
• Demonstrated a low error rate (0.5%) and rapid computation time (0.25s), indicating clinical applicability.

Conclusions:

• The proposed GECR2ANet+BBOA framework offers a significant advancement in EEG-based Parkinson's disease recognition.
• The model's high accuracy and efficiency surpass existing methods, promising improved early detection and diagnosis of Parkinson's disease.
• This approach holds potential for real-world clinical applications, enhancing diagnostic precision for neurological disorders.