Assessing dynamic functional connectivity in heterogeneous samples

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Computational Neuroscience

Background:

• Dynamic functional connectivity (dFC) analysis in fMRI aims to capture temporal brain activity patterns.
• Sliding-window methods are common for dFC estimation, but can be sensitive to data heterogeneity.
• Group comparisons of dFC, e.g., age-related differences, are a key research objective.

Purpose of the Study:

• To investigate how individual differences in non-dynamic fMRI data features affect dFC estimation.
• To determine if common dFC analysis methods can produce spurious group differences.
• To highlight the need for careful consideration of confounding factors in dFC research.

Main Methods:

• Utilized a generic simulation framework for functional magnetic resonance imaging (fMRI) data.
• Introduced variations in parameters like hemodynamic response function (HRF) shape, neural autocorrelation, connectivity strength, and measurement noise.
• Applied standard dFC analysis techniques (e.g., k-means, multilayer modularity) to simulated data.

Main Results:

• Simulated individual differences in HRF shape, noise, and other factors led to apparent dFC variations, even without true dFC changes.
• Common dFC methods detected spurious group differences when hyperparameters were not appropriately set.
• The study demonstrated that non-dynamic individual differences can mimic dynamic connectivity changes.

Conclusions:

• Individual variations in fMRI data characteristics (e.g., HRF, noise) can confound dFC group comparisons.
• Standard dFC methods may erroneously identify group differences due to data heterogeneity.
• Researchers must address potential sources of individual variability before concluding on group-specific dFC alterations.