Enhanced Graph Convolutional Network with Chebyshev Spectral Graph and Graph Attention for Autism Spectrum Disorder Classification

TL;DR

The paper tackles ASD diagnosis challenges by leveraging a multi-branch Enhanced Graph Convolutional Network that fuses multimodal ABIDE I data (rs-fMRI, sMRI, and phenotypes) through Chebyshev spectral graph convolutions and Graph Attention Networks. It introduces a population graph based on site information and modality-specific branches to capture inter-subject relationships and complex feature interactions. The model achieves 74.81% accuracy and an AUC of 0.82 on the full ABIDE I dataset, with ablation studies confirming the value of graph attention and hyperparameter tuning, and it outperforms several state-of-the-art baselines on complete data. This approach demonstrates the effectiveness of graph-based, multimodal learning for cross-site ASD classification and offers a scalable framework for integrating diverse neuroimaging and phenotypic data in clinical contexts.

Abstract

ASD is a complicated neurodevelopmental disorder marked by variation in symptom presentation and neurological underpinnings, making early and objective diagnosis extremely problematic. This paper presents a Graph Convolutional Network (GCN) model, incorporating Chebyshev Spectral Graph Convolution and Graph Attention Networks (GAT), to increase the classification accuracy of ASD utilizing multimodal neuroimaging and phenotypic data. Leveraging the ABIDE I dataset, which contains resting-state functional MRI (rs-fMRI), structural MRI (sMRI), and phenotypic variables from 870 patients, the model leverages a multi-branch architecture that processes each modality individually before merging them via concatenation. Graph structure is encoded using site-based similarity to generate a population graph, which helps in understanding relationship connections across individuals. Chebyshev polynomial filters provide localized spectral learning with lower computational complexity, whereas GAT layers increase node representations by attention-weighted aggregation of surrounding information. The proposed model is trained using stratified five-fold cross-validation with a total input dimension of 5,206 features per individual. Extensive trials demonstrate the enhanced model's superiority, achieving a test accuracy of 74.82\% and an AUC of 0.82 on the entire dataset, surpassing multiple state-of-the-art baselines, including conventional GCNs, autoencoder-based deep neural networks, and multimodal CNNs.

Figures (2)

• Figure 1: The proposed enhanced GCN architecture with Graph Attention
• Figure 2: Testing AUC curve