Rethinking Measures of Functional Connectivity via Feature Extraction

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Brain Connectivity

Background:

• Functional magnetic resonance imaging (fMRI) is a key tool for studying brain function.
• Functional connectivity (FC) quantifies relationships between brain regions using blood-oxygen-level-dependent (BOLD) signals.
• Pearson's correlation is the conventional method for FC but captures only linear dependencies.

Purpose of the Study:

• To evaluate the sufficiency of Pearson's correlation for characterizing FC.
• To identify and assess alternative FC measures.
• To understand the implications of using different FC measures on brain network analysis and population variability.

Main Methods:

• Analysis of fMRI data from a healthy adult population.
• Comparison of Pearson's correlation with eight alternative FC quantification methods.
• Assessment of FC measure consistency across task and resting-state conditions.
• Evaluation of FC measures' ability to classify age and associate with behavioral outcomes.

Main Results:

• Pearson's correlation inadequately captures complex BOLD signal inter-dependencies.
• Eight alternative FC measures demonstrated consistency between task and resting-state fMRI.
• Alternative FC measures enhanced age-based classification and behavioral outcome associations.
• Canonical large-scale brain networks appear to be dependent on the chosen FC measure.

Conclusions:

• A comprehensive, multi-metric definition of FC is more appropriate than Pearson's correlation alone.
• The choice of FC measure significantly influences the characterization of brain networks.
• Improved FC quantification promises a deeper understanding of individual differences in brain function.