Ranking Regions, Edges and Classifying Tasks in Functional Brain Graphs by Sub-Graph Entropy.

Area of Science:

• Neuroscience
• Information Theory
• Machine Learning

Background:

• Functional magnetic resonance imaging (fMRI) generates time-series data representing brain activity.
• Human brain networks are modeled as graphs where nodes are regions and edges represent correlations.
• Analyzing these complex networks is crucial for understanding brain function during various tasks.

Purpose of the Study:

• To introduce a novel information-theoretic metric, sub-graph entropy, for analyzing functional brain networks.
• To utilize node and edge entropies for ranking significant brain regions and connections.
• To classify task-based brain activity using machine learning models based on these entropy features.

Main Methods:

• Construction of brain networks from fMRI time-series data.
• Introduction and application of sub-graph entropy, node entropy, and edge entropy.
• Calculation of differential node and edge entropies to identify task-specific features.
• Development of Support Vector Machine (SVM) classifiers with a radial basis function (RBF) kernel.

Main Results:

• Sub-graph entropy effectively ranks nodes and edges in functional brain networks.
• Differential node and edge entropies highlight regions and connections associated with specific tasks (emotion, gambling).
• SVM classifiers using node entropies achieved accuracies of 0.96-0.98.
• SVM classifiers using edge entropies achieved accuracies of 0.91-0.96.

Conclusions:

• Sub-graph entropy is a powerful tool for analyzing functional brain networks.
• The proposed method accurately classifies task-related brain states using fMRI data.
• This approach offers a promising avenue for understanding brain function and developing diagnostic tools.