Interpretable Spatio-Temporal Embedding for Brain Structural-Effective Network with Ordinary Differential Equation.

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Magnetic Resonance Imaging (MRI) brain networks are crucial for understanding brain structure, function, disease, and development.
• Current functional MRI (fMRI) methods often analyze synchronous signals, limiting the capture of directional influences and temporal dynamics between brain regions.

Purpose of the Study:

• To develop an advanced framework for analyzing brain networks that captures directional influences and spatio-temporal dynamics.
• To introduce an interpretable graph learning approach for modeling the interplay between structural and effective brain networks.

Main Methods:

• Construction of brain-effective networks using the dynamic causal model.
• Introduction of the Spatio-Temporal Embedding ODE (STE-ODE) framework, incorporating directed node embedding layers.
• Modeling of spatial-temporal brain dynamics using an ordinary differential equation (ODE) model to capture network interplay.

Main Results:

• Validation of the STE-ODE framework on clinical phenotype prediction tasks using the HCP and OASIS datasets.
• Demonstration of superior performance of the proposed model compared to several state-of-the-art methods in prediction tasks.

Conclusions:

• The STE-ODE framework effectively captures dynamic inter-play between structural and effective brain networks.
• This novel approach enhances the analysis of brain connectivity for clinical applications, offering improved predictive power.