Graph-Based Deep Learning for Brain Network Analysis and Connectivity Mapping

This paper introduces graph-based deep learning for functional connectivity mapping and brain network analysis that improves accuracy, reliability, and efficiency. The method creates a link network from functional MRI data, with nodes indicating brain areas and lines exhibiting statistical connectivity. The model strengthens neural linkages through spectrum decomposition, graph convolution layers, and attention processes to improve categorization. It outperforms CNNs (89.2%), RNNs (85.6%), SVM (80.4%), and RF (78.1%) with 95.8% accuracy. It is fast (120 ms), scalable, and easy to grasp (9/10), making it ideal for large-scale scanning. The approach also detects abnormalities in neuron joining, which may aid neurological disorder diagnosis. Precision (94.5%), recall (96.2%), and F1-score (95.3%) demonstrate its dependability. It utilizes less memory than CNNs (400 MB) and RNNs (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{4 5 0 ~ M B}$</tex>) at <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{2 5 0 ~ M B}$</tex>. This study shows that graph-based deep learning works effectively in functional connectivity research. They help us understand brain-computer interfaces, cognitive function, and neurological diseases.