Tri-Clustering Dynamic Functional Network Connectivity Identifies Significant Schizophrenia Effects Across Multiple...

Area of Science:

• Neuroscience
• Computational Biology
• Medical Imaging Analysis

Background:

• Individual brain imaging data exhibit unique variations often overlooked by group-average or supervised prediction methods.
• Current dynamic functional network connectivity (dFNC) analyses using sliding windows may miss subtle connectivity patterns due to data homogeneity issues.
• Existing methods like k-means clustering on dFNC windows may fail to identify connectivity signatures spanning smaller subsets of neural components.

Purpose of the Study:

• To introduce and validate Dynamic-N-way tri-clustering (dNTiC) for analyzing high-dimensional brain imaging data, accounting for individual variability.
• To compare brain connectivity patterns between individuals with schizophrenia (SZ) and healthy controls (HC) using the dNTiC method.
• To enhance the interpretability and sensitivity of dFNC measurements for heterogeneous disorders like schizophrenia.

Main Methods:

• Developed Dynamic-N-way tri-clustering (dNTiC), a novel approach that incorporates a homogeneity benchmark for data subgrouping.
• dNTiC sorts dFNC states by maximizing inter-individual similarity and minimizing intra-state component variance.
• Applied dNTiC to resting-state functional magnetic resonance imaging (fMRI) data from SZ and HC groups to analyze dynamic connectivity states.

Main Results:

• dNTiC identified significant differences in dFNC states between SZ and HC groups across distinct brain regions.
• Schizophrenia (SZ) subjects exhibited hypoconnectivity in subcortical and default mode networks, and hyperconnectivity in sensory networks compared to healthy controls (HC).
• Significant differences were observed in the recurrence time of specific dFNC states between SZ and HC groups, with HC showing stronger sensory network connectivity.

Conclusions:

• The proposed dNTiC method effectively leverages individual data variance, offering enhanced interpretability and sensitivity for analyzing complex brain disorders.
• dNTiC provides a more nuanced understanding of dynamic functional connectivity alterations in schizophrenia compared to traditional methods.
• The findings underscore the utility of dNTiC for characterizing transient and complex neural patterns in high-dimensional neuroimaging studies.