Characterising brain network topologies: A dynamic analysis approach using heat kernels.

Area of Science:

• Neuroscience
• Network Science
• Computational Biology

Background:

• The human brain is increasingly modeled as a complex network to understand its organization.
• Dynamic brain connectivity is crucial for information transport, yet challenging to quantify.
• Existing methods may not fully capture the complex dynamics of brain networks.

Purpose of the Study:

• To propose a novel network modeling approach using the heat kernel for analyzing brain connectivity.
• To define new features quantifying heat propagation changes in structural brain networks.
• To assess the utility of these features in classifying networks and predicting clinical outcomes.

Main Methods:

• Applied a heat kernel-based network modeling approach to structural brain networks.
• Defined features that quantify changes in heat propagation dynamics.
• Validated features on synthetic networks with varying topologies and a clinical cohort of preterm infants.

Main Results:

• Heat kernel features effectively capture network efficiency and topological properties, such as small-world architecture.
• The features demonstrated discriminative power between different network topologies.
• In preterm infants, heat kernel features accurately predicted motor function at two years of age (82.3% accuracy).

Conclusions:

• The heat kernel approach offers a powerful new metric for quantifying brain network organization and efficiency.
• These features have significant potential for characterizing brain development and the impact of disease.
• The methodology shows promise for clinical applications, particularly in predicting neurodevelopmental outcomes.