ASD-HybridNet: A hybrid deep learning framework for detection of autism spectrum disorder

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Autism spectrum disorder (ASD) diagnosis relies on subjective behavioral assessments, limiting accuracy and early detection.
• Functional magnetic resonance imaging (fMRI) offers objective biomarkers for neurological conditions.

Purpose of the Study:

• To develop and validate a hybrid deep learning framework (ASD-HybridNet) for improved autism spectrum disorder detection.
• To leverage multimodal fMRI data for more accurate and earlier ASD diagnosis.

Main Methods:

• Integration of region of interest (ROI) time series data and functional connectivity (FC) maps from fMRI.
• Development of a hybrid deep learning architecture, ASD-HybridNet.
• Validation using the Autism Brain Imaging Data Exchange (ABIDE) dataset.

Main Results:

• ASD-HybridNet demonstrated superior performance in ASD detection compared to existing methods.
• The integration of ROI and FC data significantly improved classification accuracy.
• The framework shows potential for enhancing early and accurate ASD diagnosis.

Conclusions:

• The proposed ASD-HybridNet framework offers a promising, data-driven approach for autism spectrum disorder diagnosis.
• Multimodal fMRI data integration via deep learning can overcome limitations of traditional diagnostic methods.