Comparing functional connectivity matrices: A geometry-aware approach applied to participant identification.

Area of Science:

• Neuroscience
• Network Science
• Data Analysis

Background:

• Functional connectivity matrices capture brain network organization.
• Current methods like Euclidean distance and Pearson correlation have limitations in analyzing these matrices.
• Participant identification (fingerprinting) relies on comparing these matrices.

Purpose of the Study:

• To introduce and evaluate a novel geodesic distance metric for functional correlation matrices.
• To assess the effectiveness of this metric for participant identification.
• To explore the geometry of brain functional connectivity.

Main Methods:

• Proposed a geodesic distance metric accounting for the non-Euclidean geometry of correlation matrices.
• Evaluated the metric using participant identification (fingerprinting) with resting-state and task-based fMRI data.
• Compared performance against Pearson correlation and analyzed subnetworks.

Main Results:

• Geodesic distance achieved over 95% accuracy in resting-state participant identification.
• Accuracy improved by 2-20% on various tasks compared to Pearson correlation.
• Identification accuracy increased by over 10% using specific subnetwork pairs.
• The geodesic approach outperformed Pearson correlation even with less data.

Conclusions:

• Geodesic distance is a powerful tool for analyzing brain functional connectivity matrices.
• This metric enhances participant identification accuracy and provides insights into brain network geometry.
• The approach offers a more effective way to compare brain correlation structures across studies.