A complex-valued functional brain connectivity descriptor amenable to Riemannian geometry.

Area of Science:

• Neuroscience
• Signal Processing
• Machine Learning

Background:

• Functional connectivity analysis is crucial for understanding brain networks.
• Traditional methods often struggle with high-dimensional data and capturing temporal dynamics.
• Phase synchronization and time-lag information are key but challenging to integrate.

Purpose of the Study:

• To introduce a novel phase-based functional connectivity descriptor.
• To leverage Riemannian geometry for analyzing complex brain network data.
• To improve the differentiation of brain network organization under different conditions.

Main Methods:

• Developed a novel phase-based functional connectivity descriptor using complex-valued measurements.
• Utilized Hermitian Positive Definite (HPD) matrices to represent brain networks.
• Adapted Riemannian geometry-based dimensionality reduction for discriminative analysis.

Main Results:

• The proposed descriptor captures both synchronization strength and time-lag of neural oscillations.
• Dimensionality reduction on HPD matrices effectively identifies key brain regions.
• Validated framework shows high classification performance for attentive vs. passive visual responses using EEG data.

Conclusions:

• The novel descriptor offers a powerful tool for brain network analysis.
• This approach integrates phase synchronization, time-lag, and Riemannian geometry.
• The method facilitates neuroscientific exploration and improves classification accuracy.