Disorder-specific neurodynamic features in schizophrenia inferred by neurodynamic embedded contrastive variational...

Area of Science:

• Neuroscience
• Computational Psychiatry
• Systems Biology

Background:

• Schizophrenia (SCZ) pathophysiology remains complex, with micro-level neural alterations impacting macro-scale brain function.
• Neurodynamic models offer a framework to understand how these changes propagate and affect neural circuits.

Purpose of the Study:

• To integrate a neurodynamic model with a Contrastive Variational Autoencoder (CVAE) for SCZ feature extraction.
• To evaluate macro-scale, SCZ-specific features, including subject-level, region-level, and time-varying states.
• To investigate the relationship between these features and SCZ symptoms.

Main Methods:

• Integration of a neurodynamic model with CVAE for feature extraction.
• Representational similarity analysis and deep learning classification to confirm feature specificity.
• Analysis of neurodynamic system attractor characteristics.
• Partial Least Squares (PLS) regression to map features to symptoms.

Main Results:

• Robust model fitting demonstrated across a multi-site dataset.
• SCZ-specific features confirmed to capture disorder-related information.
• Distinct attractor space patterns identified between SCZ-specific and shared neural states.
• Two sets of correlated modes linking SCZ features to symptoms (negative/general and positive) were identified.

Conclusions:

• The study successfully extracted and validated SCZ-specific neurodynamic features and states.
• Findings highlight distinct attractor dynamics in schizophrenia.
• Identified feature-symptom correlations suggest unique underlying molecular mechanisms for different symptom clusters.
• This work provides a foundation for understanding SCZ pathophysiology through neurodynamic modeling.