Geodesic Distance on Optimally Regularized Functional Connectomes Uncovers Individual Fingerprints.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Brain Connectivity Analysis

Background:

• Functional connectomes (FCs) serve as unique individual fingerprints, crucial for personalized medicine in neurological and psychiatric disorders.
• Current FC comparison methods, like Pearson's correlation, may not fully capture the non-Euclidean geometry of brain data.
• Geodesic distance offers a more accurate comparison but requires positive-definite matrices, often necessitating regularization.

Purpose of the Study:

• To investigate the role and impact of regularization in enhancing functional connectome (FC) fingerprinting using geodesic distance.
• To determine if regularization is merely an algebraic necessity or a method to improve individual identification accuracy.
• To establish optimal regularization strategies for maximizing the reliability and differentiability of FC comparisons.

Main Methods:

• Exploration of geodesic distance for comparing functional connectomes (FCs).
• Systematic analysis of the effect of varying regularization magnitudes on FC fingerprint reproducibility and distinctiveness.
• Investigation of data-set dependency of optimal regularization across different conditions, parcellations, and scanning parameters.

Main Results:

• Regularization is not just for invertibility but can significantly enhance individual fingerprinting accuracy.
• An optimal regularization magnitude exists and is dependent on dataset specifics (condition, parcellation, scan length, frame count).
• Fixed, universal regularization strategies underperform compared to data-set specific optimal regularization.

Conclusions:

• Optimal regularization is essential for maximizing the potential of geodesic distance in individual brain fingerprinting.
• Estimating optimal regularization for each dataset is critical for reliable within-subject reproducibility and between-subject differentiability.
• Pairwise geodesic distances at optimal regularization provide a robust measure for quantifying individual differences in functional brain organization.