Investigating statistical differences in connectivity patterns properties at single subject level: a new resampling...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Network Science

Background:

• Multivariate autoregressive (MVAR) models are standard for estimating effective brain connectivity.
• Current MVAR methods lack a single-subject distribution, hindering within-subject statistical comparisons across conditions.
• This limitation is particularly problematic for heterogeneous clinical populations.

Purpose of the Study:

• To develop a novel method for generating a distribution of brain connectivity values within a single subject.
• To enable robust statistical assessment of changes in brain network properties for individual patients.
• To address the limitations of current effective connectivity estimation techniques.

Main Methods:

• Proposed a novel approach based on small perturbations of network properties to construct a connectivity distribution.
• Utilized graph theory to derive brain connectivity indexes.
• Validated the method through a simulation study and application to real electroencephalography (EEG) data.

Main Results:

• The novel approach successfully generated a distribution of connectivity values for single subjects.
• The method allowed for the assessment of significant changes in graph theory-derived brain connectivity indexes.
• Feasibility and applicability were demonstrated in both simulated and real EEG data.

Conclusions:

• The proposed method provides a viable solution for within-subject statistical comparisons of brain connectivity.
• This advancement is significant for analyzing brain network dynamics in clinical populations with heterogeneous conditions.
• The approach enhances the utility of MVAR methods for personalized neuroscience research.