An enriched granger causal model allowing variable static anatomical constraints.

Area of Science:

• Neuroscience
• Computational Psychiatry
• Medical Imaging Analysis

Background:

• Functional brain connectivity is influenced but not solely determined by anatomical connections, particularly in dynamic contexts.
• Understanding dynamic effective connectivity is crucial for distinguishing depression and exploring its pathogenesis.

Purpose of the Study:

• To propose an enriched Granger causal model (GCM) integrating anatomical prior information to describe dynamic effective connectivity.
• To differentiate depression and investigate its underlying mechanisms using this novel approach.

Main Methods:

• Anatomical information was transformed and integrated into a normal GCM using a variational Bayesian model.
• Magnetoencephalography (MEG) and diffusion tensor imaging (DTI) data from 24 depressive patients and 24 controls were analyzed.
• Sliding windowed MEG signals under sad facial stimuli were used with the enriched GCM to calculate dynamic effective connectivity, followed by feature selection and classification.

Main Results:

• The enriched GCM with anatomical priors demonstrated superior performance over the normal GCM in terms of model errors and recognition accuracy.
• Support Vector Machine (SVM) classification achieved 85.42% accuracy (87.50% sensitivity, 83.33% specificity) for effective connectivity with anatomical priors.
• Discriminative feature analysis indicated that the enriched GCM better identified more stringent and less dynamic brain function in depression.

Conclusions:

• The proposed enriched GCM is valuable for exploring dynamic functional dysfunction in depression.
• This approach shows promise for improving depression recognition and understanding its neurobiological underpinnings.