NF-GAT: A Node Feature-Based Graph Attention Network for ASD Classification.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Autism spectrum disorder (ASD) presents diagnostic challenges.
• Functional connectivity (FC) analysis using fMRI shows promise in understanding brain differences in ASD.

Purpose of the Study:

• To develop and evaluate a graph attention network for recognizing autism spectrum disorders (ASD).
• To leverage functional connectivity (FC) features derived from fMRI data for improved ASD diagnosis.

Main Methods:

• A novel Node Features Graph Attention Network (NF-GAT) was proposed.
• Subject data were modeled as graphs with node features derived from fMRI.
• Graph attention layers were employed to learn discriminative node information for classification.

Main Results:

• The NF-GAT model demonstrated significant advantages over existing methods in ASD classification.
• The proposed NF-GAT achieved superior classification performance.

Conclusions:

• The NF-GAT model is effective for autism spectrum disorder classification.
• This approach offers a promising tool for objective ASD diagnosis using neuroimaging data.