Intermediately synchronised brain states optimise trade-off between subject specificity and predictive capacity.

Area of Science:

• Neuroscience
• Cognitive Science
• Data Science

Background:

• Functional connectivity (FC) quantifies statistical dependencies between brain areas.
• Temporal fluctuations in FC are studied using edge time series (ETS).
• High-amplitude co-fluctuation (HACF) time points are thought to drive FC and individual differences.

Purpose of the Study:

• To systematically evaluate the contribution of different FC time points to brain-behavior associations.
• To assess the predictive utility of FC estimates across varying co-fluctuation levels.
• To leverage machine learning (ML) for analyzing temporal dynamics in functional magnetic resonance imaging (fMRI).

Main Methods:

• Computation of edge time series (ETS) from fMRI data.
• Categorization of time points based on co-fluctuation amplitude levels.
• Application of machine learning (ML) models to assess predictive capacity.
• Evaluation of subject specificity and phenotype prediction using different FC time points.

Main Results:

• Time points with lower and intermediate co-fluctuation levels demonstrated the highest subject specificity.
• These lower and intermediate co-fluctuation time points also exhibited the greatest predictive capacity for individual phenotypes.
• The predictive utility of FC estimates varies significantly across different co-fluctuation levels.

Conclusions:

• Brain-behavior associations are not solely driven by high-amplitude co-fluctuation events.
• Lower and intermediate amplitude time points in functional connectivity offer crucial insights into individual differences.
• Rethinking the focus on high-amplitude events is necessary for a comprehensive understanding of brain-behavior relationships.