Extracting default mode network based on graph neural network for resting state fMRI study.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Functional magnetic resonance imaging (fMRI) is crucial for understanding brain function in health and disease.
• Analyzing functional brain connectivity, particularly the default mode network (DMN), is key to explaining various conditions.
• Existing methods for DMN extraction have limitations in robustness and assumptions.

Purpose of the Study:

• To introduce and evaluate graphSAGE, a graph neural network, for analyzing resting-state fMRI (rs-fMRI) data.
• To extract the default mode network (DMN) using graphSAGE.
• To compare graphSAGE's performance against established methods for DMN identification.

Main Methods:

• Utilized graphSAGE, a deep learning graph neural network, for rs-fMRI data analysis.
• Applied graphSAGE to extract the default mode network (DMN).
• Compared graphSAGE with seed-based correlation, independent component analysis, and dictionary learning using real fMRI data.

Main Results:

• GraphSAGE demonstrated superior robustness and reliability in DMN extraction compared to traditional methods.
• The graphSAGE approach defined clearer regions of interest for the DMN.
• GraphSAGE requires fewer and more relaxed assumptions, enabling simultaneous single and group subject analysis.

Conclusions:

• GraphSAGE is a powerful and effective deep learning tool for analyzing brain functional connectivity from rs-fMRI data.
• This method provides a more robust and reliable approach to DMN identification.
• The flexibility of graphSAGE in handling both individual and group analyses offers significant advantages.