Decoding Task-Specific Cognitive States with Slow, Directed Functional Networks in the Human Brain.

Area of Science:

• Neuroscience
• Cognitive Science
• Data Science

Background:

• Brain region interactions are key to cognitive functions.
• Functional magnetic resonance imaging (fMRI) measures these interactions via functional connectivity.
• Lag-based functional connectivity estimates from fMRI are often considered unreliable due to slow hemodynamic responses.

Purpose of the Study:

• To investigate if Granger-Causality (GC) networks from fMRI data contain useful information for classifying cognitive states.
• To compare the utility of instantaneous and lag-based GC functional connectivity.
• To explore the relationship between GC connectivity and cognitive performance.

Main Methods:

• Estimated instantaneous and lag-based GC functional connectivity networks from fMRI data of 1000 participants (Human Connectome Project).
• Used a linear classifier to discriminate between seven task and resting brain states.
• Employed network simulations to understand timescale exploitation by GC methods.
• Correlated GC connectivity variations with inter-individual cognitive scores.

Main Results:

• GC networks from fMRI reliably discriminated between cognitive states (>80% accuracy).
• Instantaneous and lag-based GC methods captured interactions at fast and slow timescales, respectively.
• These two methods identified complementary, task-core networks in human fMRI data.
• GC connectivity variations explained individual differences in cognitive scores.

Conclusions:

• Despite caveats, GC networks estimated from fMRI contain valuable information for cognitive state classification.
• Instantaneous and lag-based fMRI connectivity methods reveal complementary aspects of brain function.
• Slow, directed functional interactions estimated via fMRI can serve as useful markers for behaviorally relevant cognitive states.