Spectral GNN for fMRI Cognitive Task Classification

This paper introduces a novel Spectral Graph Neural Network (SpectralBrainGNN) for classifying cognitive tasks using fMRI data. The approach leverages graph neural networks to model brain connectivity, capturing complex topological dependencies. The high classification accuracy (96.25%) on the HCPTask dataset and the public availability of the implementation are significant contributions, promoting reproducibility and further research in neuroimaging and machine learning.

• •Proposes SpectralBrainGNN, a spectral convolution framework for cognitive task classification.
• •Utilizes graph neural networks to model brain connectivity from fMRI data.
• •Achieves high classification accuracy on the HCPTask dataset.
• •Provides publicly available implementation for reproducibility.