A Survey of Methods and Architectures for Parkinson's Disease Recognition from EEG Using Attention-Based Sparse Graph Convolutional Neural Networks

Abstract

Parkinson’s disease (PD) is a progressive neurodegenerative disorder
that significantly affects motor and cognitive functions, necessitating
early and accurate diagnosis. Electroencephalography (EEG) has
emerged as a promising non-invasive tool for detecting neural
abnormalities associated with PD. Recent advancements in artificial
intelligence, particularly deep learning, have revolutionized EEG-based
PD recognition. Among these, attention-based sparse graph
convolutional neural networks (ASGCNN) have demonstrated superior
capability by modeling functional brain connectivity and focusing on
informative EEG channels. This survey reviews recent methods in EEG
based PD detection, emphasizing graph neural networks (GNN),
attention mechanisms, and hybrid deep learning architectures. GNN
models effectively capture spatial relationships between EEG channels,
while attention mechanisms enhance interpretability and feature
selection. The incorporation of sparsity constraints further improves
efficiency by eliminating redundant connections. The ASGCNN
framework models channel relationships using graphs, applies attention
for channel selection, and uses sparsity to reduce redundancy,
significantly improving classification performance. The paper also
discusses challenges such as data variability, computational complexity,
and lack of explainability. Future directions include multimodal learning,
lightweight architectures, and explainable AI for clinical adoption.