Diagnosis of Autism Spectrum Disorder (ASD) Using Recursive Feature Elimination-Graph Neural Network (RFE-GNN) and...

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Autism spectrum disorder (ASD) is a complex neurodevelopmental condition affecting social, behavioral, and communication skills.
• The exact causes of ASD are not fully understood, but brain activity patterns are implicated.
• Current diagnostic methods can be improved with more objective, data-driven approaches.

Purpose of the Study:

• To propose a novel framework for autism spectrum disorder (ASD) detection.
• To integrate functional magnetic resonance imaging (fMRI) and phenotypic data for enhanced diagnostic accuracy.
• To develop a computationally efficient and effective ASD diagnostic tool.

Main Methods:

• Utilized recursive feature elimination (RFE) for selecting key features from fMRI data.
• Employed graph neural networks (GNN) for extracting informative features from selected fMRI data.
• Developed a phenotypic feature extractor (PFE) and synergistically fused extracted features.

Main Results:

• The proposed framework achieved 78.7% and 80.6% accuracy on the ABIDE dataset.
• Demonstrated competitive performance compared to existing state-of-the-art ASD detection methods.
• Validated the efficacy of integrating fMRI and phenotypic data for ASD identification.

Conclusions:

• The novel framework offers a promising direction for developing effective diagnostic tools for ASD.
• Combining fMRI and phenotypic data through GNN and RFE enhances ASD detection capabilities.
• This approach contributes to advancing objective and accurate methods for ASD diagnosis.