Functional connectivity change as shared signal dynamics.

Area of Science:

• Neuroscience
• Data Analysis

Background:

• Functional connectivity analysis is crucial for understanding brain function by examining inter-region interactions.
• Current methods often use variance normalization, which can be misled by independent signal fluctuations, obscuring true connectivity changes.
• This sensitivity to unshared signal reduces the accuracy of functional connectivity estimates like correlations.

Purpose of the Study:

• To investigate alternative functional connectivity measures that are less sensitive to independent signal fluctuations.
• To improve the interpretability and accuracy of functional connectivity change detection.
• To introduce and evaluate covariance and covariance conjunction as improved methods.

Main Methods:

• Simulations were used to compare functional connectivity measures with and without variance normalization.
• Covariance, a non-normalized correlation, was explored for its ability to isolate shared signal.
• A novel "covariance conjunction" method was developed, combining normalized and non-normalized approaches.

Main Results:

• Simulations demonstrated that covariance effectively isolates differences in shared signal, enhancing interpretability.
• Covariance and covariance conjunction methods successfully detected functional connectivity changes in diverse functional MRI datasets (task, rest, clinical, non-clinical).
• The choice between correlation, covariance, and covariance conjunction significantly impacts observed functional connectivity results.

Conclusions:

• Forgoing variance normalization and utilizing covariance or covariance conjunction improves the detection and interpretation of functional connectivity changes.
• These methods offer practical and theoretical advantages for analyzing brain connectivity in various contexts.
• Isolating shared signal changes provides a more accurate representation of inter-region interaction.