Whole-Brain Modelling and Whole-Brain Dynamical Analysis in ALZHEIMER’S DISEASE

@Centre for Brain and Cognition, Computational Neuroscience Group, Department of Information and Communication Technologies, Universitat Pompeu Fabra, Barcelona, Spain (Nov 2024 - June 2025)

Supervised by - Prof. Gustavo Deco and Dr. Jakub Vohryzek

Alzheimer’s disease (AD) progressively disrupts the brain’s ability to sustain its intrinsic functional organization, yet the underlying dynamical mechanisms remain poorly understood. Beyond structural degeneration, recent studies suggest that AD reflects a breakdown in the non-equilibrium nature of brain dynamics, as a result collapse in the temporal asymmetry occurs that supports hierarchical information processing. In this study, we quantify this loss of irreversibility in empirical fMRI time series and examine its mechanistic origin using a generative whole-brain modeling framework based on coupled Hopf oscillators. By fitting both functional connectivity and arrow of time asymmetry, we extract individualized generative effective connectivity (GEC) profiles that reflect directional interactions between brain regions. Our analysis reveals a systematic degradation of temporal asymmetry in AD , particularly within the default mode, salience, and limbic networks. As the disease progresses, frontal hubs lose their driver status, while posterior cortices gain dominant outflow, suggesting a posterior shift in the causal hierarchy. These changes are not captured by symmetric functional connectivity alone and emerge only when irreversibility constraints are incorporated during model optimization. The resulting GEC matrices also reveal spatially distinct asymmetry and trophic coherence patterns, offering a refined lens on hierarchical collapse in AD. This integrated framework bridges thermodynamic theory, whole-brain modeling, and empirical data to uncover signatures of altered information flow that precede overt cognitive decline. Our findings provide a principled foundation for developing brain-state aware diagnostic tools and temporally informed intervention strategies.