Kernel-Transformed Functional Connectivity Entropy Reveals Network Dedifferentiation in Bipolar Disorder

Area of Science:

• Neuroscience
• Psychiatry
• Data Science

Background:

• Resting-state functional MRI (rs-fMRI) commonly uses linear correlation, potentially missing network distributional features.
• Bipolar disorder (BD) pathophysiology may involve altered functional brain networks.

Purpose of the Study:

• Introduce a kernel-transformed functional connectivity (FC) entropy framework.
• Quantify network dedifferentiation in bipolar disorder (BD).
• Explore FC entropy as a potential biomarker for BD.

Main Methods:

• Applied a Gaussian kernel function for nonlinear similarity transformation (reweighting) of linear correlation matrices.
• Calculated multiscale entropy (global, modular, nodal) to assess connectivity weight distribution uniformity.
• Controlled for head motion (Mean FD) and analyzed correlations with manic symptom severity (YMRS).

Main Results:

• Bipolar disorder (BD) patients exhibited significantly higher global, Default Mode, Salience, and Somatosensory-Motor network entropy compared to Normal Controls (NCs).
• This indicates widespread network dedifferentiation (distributional flattening) in BD.
• Higher global entropy correlated negatively with manic symptom severity (YMRS).

Conclusions:

• Kernel-transformed FC entropy is a distribution-sensitive metric complementing linear approaches.
• Network dedifferentiation is a key pathophysiological feature in bipolar disorder (BD).
• This framework shows promise for characterizing neural network dysregulation in BD.