UNRAVELING BRAIN NETWORK DYNAMICS: A NOVEL APPROACH THROUGH AN ENHANCED NEUROLIB FRAMEWORK

Abstract

Understanding brain dynamics is crucial for advancing both basic neuroscience and clinical research. In particular, resting-state brain activity offers valuable insights into how different regions of the brain interact in a spontaneous, baseline state. Examining these interactions opens the possibility to obtain a deeper understanding of the brain¿s functional connectivity, which is essential for studying normal brain function as well as various neurological disorders.

Recent advancements in computational neuromodelling have made significant advances in this area, with tools like The Virtual Brain and Neurolib providing powerful frameworks for simulating brain activity. These tools integrate empirical data derived from various medical imaging techniques, including fMRI, EEG, MEG, DWI, and DTI, to create accurate models of brain networks. This allows researchers to simulate brain activity and better understand it.

In this thesis, Neurolib is implemented as part of the methodology. Several improvements were made to its framework, notably the introduction of Heun's integration method, which provided better results compared to the traditional Euler method, and the development of a custom multimodel framework that proved nearly 600 times faster than the existing implementation. Additionally, the geodesic distance is introduced as a metric for evaluating model performance, allowing for more meaningful comparisons between empirical and simulated functional connectivity matrices, overcoming the limitations of traditional correlation methods.

Furthermore, a novel phenomenological model based on a subcritical Hopf bifurcation was proposed for the thalamus and integrated into a corticothalamic framework. This approach demonstrated that, while biophysical models like ALN-Costa offer more accurate representations of brain dynamics, Hopf models still hold significant potential for capturing oscillatory behavior. despite limitations in simulating BOLD signals.

The thesis also explored the propagation of oscillatory activity within Hopf-based networks. The results showed that oscillations propagate from oscillatory nodes to stable nodes through coupling. The strength of coupling was found to be a more decisive factor than the distance or delay between nodes.

With the promising results obtained, this research contributes to the advancement in computational modelling tools, providing a foundation for future research.