Few temporally distributed brain connectivity states predict human cognitive abilities

Area of Science:

• Neuroscience
• Cognitive Neuroscience
• Brain Imaging

Background:

• Human functional brain connectivity exhibits temporal fluctuations, including states of high and low cofluctuation.
• Rare, high cofluctuation states are subject-specific and reflect intrinsic functional network architecture.
• The contribution of these network-defining states to individual cognitive abilities remains unclear.

Purpose of the Study:

• To investigate whether specific functional brain connectivity states predict individual differences in intelligence.
• To introduce and validate a novel framework, CMEP, for predicting cognitive abilities from brain connectivity data.
• To explore the temporal dynamics of brain connectivity relevant to cognitive function.

Main Methods:

• Development of the Connectivity-based Mapping of Eigenvector Patterns (CMEP) framework for analyzing resting-state fMRI data.
• Analysis of temporally separated time frames (<1.5% of total scan time) from resting-state fMRI.
• Validation of findings in two independent samples (N=263 and N=831).

Main Results:

• A small number of temporally distinct time frames (<1.5% of scan time) significantly predict individual differences in intelligence (p < .001).
• Network-defining time frames characterized by particularly high cofluctuation did not predict intelligence.
• Predictive information for intelligence is distributed across the entire brain connectivity time series, not confined to specific high-connectivity states.
• Multiple functional brain networks contribute to intelligence prediction.

Conclusions:

• Cognitive abilities are predictable from brief, temporally distributed segments of functional brain connectivity.
• High cofluctuation states, while fundamental to connectome architecture, do not solely explain individual differences in intelligence.
• Understanding cognitive abilities requires analyzing temporal dynamics across the entire brain connectivity time series, rather than focusing on isolated high-connectivity events.