How well does structural equation modeling reveal abnormal brain anatomical connections? An fMRI simulation study

Area of Science:

• Neuroscience
• Computational Neuroscience
• Medical Imaging Analysis

Background:

• Brain disorders often stem from altered anatomical connectivity between brain regions.
• Effective connectivity analysis aims to infer directed interactions between neural elements.
• Functional magnetic resonance imaging (fMRI) is a key tool for studying brain activity.

Purpose of the Study:

• To investigate if effective connectivity differences reveal anatomical alterations in patient populations.
• To assess the utility of Structural Equation Modeling (SEM) for detecting connectivity changes in simulated fMRI data.
• To compare SEM results between task conditions and between simulated normal and patient groups.

Main Methods:

• Simulated fMRI time-series data from a neurobiologically realistic network model.
• Manipulation of anatomical connectivity to simulate a 'patient' group with one weakened connection.
• Application of Structural Equation Modeling (SEM) to analyze interregional effective connectivity.

Main Results:

• A significantly reduced effective connectivity was found, corresponding to the weakened anatomical connection during a task.
• Patient networks showed reduced effective connections downstream from the lesion, but some upstream connections were larger.
• Total error variance in SEM models best distinguished patient from normal networks.

Conclusions:

• SEM can detect alterations in effective connectivity mirroring anatomical changes, even with partial network damage.
• Caution is advised when applying effective connectivity methods to fMRI data from non-normal populations.
• Observed abnormalities in functional interactions may arise even when only a portion of the neural network is compromised.