MCDGLN: Masked connection-based dynamic graph learning network for autism spectrum disorder.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Medical Imaging Analysis

Background:

• Autism Spectrum Disorder (ASD) is a complex neurodevelopmental condition.
• Previous research often overlooked dynamic brain activity and network noise.
• Static cerebral interaction analysis limits understanding of ASD's dynamic nature.

Purpose of the Study:

• To develop a novel network learning framework, the Masked Connection-based Dynamic Graph Learning Network (MCDGLN), for analyzing brain activity in ASD.
• To capture dynamic brain characteristics and refine functional connectivity by reducing network noise.
• To improve the classification accuracy between ASD and typical control groups.

Main Methods:

• Utilized sliding temporal windows on BOLD signals to capture dynamic brain characteristics.
• Employed a weighted edge aggregation (WEA) module with cross-convolution for integrating dynamic functional connectivity.
• Applied a hierarchical graph convolutional network (HGCN) with self-attention for topological feature extraction and an attention-based connection encoder (ACE) for feature refinement.
• Refined static functional connections using a task-specific mask to reduce noise and irrelevant links.

Main Results:

• The MCDGLN framework achieved 73.3% classification accuracy between ASD and Typical Control (TC) groups using the ABIDE I dataset (1035 subjects).
• The WEA and ACE modules were crucial in refining connectivity and enhancing classification accuracy.
• Demonstrated the importance of dynamic connectivity analysis and noise reduction in identifying ASD-specific neural features.

Conclusions:

• The MCDGLN framework offers a promising approach for analyzing dynamic brain connectivity in ASD.
• The study highlights the significance of dynamic functional connectivity and noise reduction in understanding ASD.
• The findings provide new insights into the neurophysiological underpinnings of Autism Spectrum Disorder.