A novel graph neural network method for Alzheimer's disease classification.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder.
• Early diagnosis of AD is crucial for effective treatment and management.
• Existing computer-aided diagnostic (CAD) methods for AD classification have limitations.

Purpose of the Study:

• To propose a novel graph neural network method, the Brain Graph Attention Network (BGAN), for Alzheimer's disease classification.
• To leverage brain graph data for modeling AD classification as a graph classification task.
• To enhance the accuracy and interpretability of AD diagnosis using advanced machine learning.

Main Methods:

• Utilized brain graph data to represent AD classification as a graph classification problem.
• Developed a novel Brain Graph Attention Network (BGAN) incorporating local and global attention layers.
• The local attention layer captures interactions between neighboring nodes, while the global attention layer determines node importance for graph representation.

Main Results:

• The BGAN model was trained and evaluated on two public Alzheimer's disease datasets.
• Experimental results demonstrated that BGAN significantly outperformed six classic models in AD classification.
• The proposed model shows potential for accurate and reliable AD diagnosis.

Conclusions:

• The Brain Graph Attention Network (BGAN) is a powerful and effective model for Alzheimer's disease classification.
• BGAN offers insights into brain regions associated with AD and its progression.
• This approach holds promise for improving early detection and understanding of Alzheimer's disease.