Artificial Intelligence Techniques for Parkinson's Disease Recognition from EEG Using Attention-Based Sparse Graph Convolutional Neural Networks: Trends and Challenges

Abstract

Parkinson’s Disease (PD) is a progressive neurodegenerative disorder that significantly affects motor and cognitive functions, making early and accurate diagnosis crucial. Electroencephalography (EEG) has emerged as a promising non-invasive tool for detecting neural abnormalities associated with PD. However, the complex, non-linear, and high-dimensional nature of EEG signals presents challenges for traditional analytical methods. Recent advancements in artificial intelligence (AI), particularly deep learning and graph neural networks (GNNs), have enabled improved modeling of EEG data by capturing spatial and functional brain connectivity.

This paper presents a comprehensive review of AI-based techniques for PD recognition using EEG, with a focus on attention-based sparse graph convolutional neural networks (ASGCNN). These models effectively represent EEG channels as graph structures, apply attention mechanisms to identify critical brain regions, and incorporate sparsity constraints to reduce noise and computational complexity. Experimental studies demonstrate that ASGCNN models achieve superior performance compared to conventional machine learning and deep learning approaches, with classification accuracies exceeding 87%.

The paper further discusses key trends, challenges, and future research directions, including hybrid architectures, explainable AI, and real-time clinical deployment. The findings highlight the transformative potential of AI-driven graph-based models in advancing EEG-based PD diagnosis.