Path analysis: A method to estimate altered pathways in time-varying graphs of neuroimaging data

Area of Science:

• Neuroscience
• Computational Psychiatry
• Network Science

Background:

• Graph-theoretical methods are established for studying brain networks in psychiatric disorders.
• Existing research often overlooks the significance of paths connecting brain regions, focusing instead on global network metrics.
• Disruptions in multistep paths within brain networks may hold crucial information for understanding disease mechanisms.

Purpose of the Study:

• To develop and apply a novel algorithm for detecting alterations in multistep brain network paths in schizophrenia.
• To identify specific edges contributing to path differences between individuals with schizophrenia and controls.
• To characterize individuals by analyzing changes in brain network paths, moving beyond traditional pairwise relationships.

Main Methods:

• Developed an algorithm to estimate edges contributing to multistep paths, comparing schizophrenia patients to controls.
• Employed a covariance decomposition method to examine shared and unique paths between groups.
• Applied the path analysis method to resting-state functional magnetic resonance imaging (fMRI) data from schizophrenia patients and healthy controls.

Main Results:

• Identified significant 'disconnectors' in schizophrenia, particularly within the default mode and cognitive control networks.
• Discovered new edges that generate additional paths in the schizophrenia group.
• Observed that while paths exist in both groups, they exhibit unique trajectories and contribute differently to network decomposition.

Conclusions:

• The proposed path analysis method offers a novel approach to characterize individuals based on brain network path alterations.
• Path-based metrics derived from neuroimaging data show promise for improving the understanding and diagnosis of psychiatric disorders like schizophrenia.
• This approach enhances the characterization of brain network disruptions by focusing on the information encoded in paths, not just direct connections.