P095 Network Complexity For Whole-Brain Dynamics Estimated From fMRI Data
Matthieu Gilson*1, Gorka Zamora-López2
1Faculty of Medecine, Aix-Marseille University, Marseille, France 2Center for Brain and Cognition, University Pompeu Fabra, Barcelona, Spain
*Email: matthieu.gilson@univ-amu.fr
Introduction
The study of complex networks has shown a fast growth in the past decades. In particular, the study of the brain as a network has benefited from the increasing availability of datasets, such as magnetic resonance imaging (MRI). This has generated invaluable insights about cognition with subjects performing tasks in the scanner, as well as alterations thereof to gain a better understanding of neuropathologies.
Methods Here we review our recent work on the estimation of effective connectivity (EC) at the whole-brain level [1]. In a nutshell, a network model can be optimized to reproduce and characterize the subject- and task-specific dynamics. This EC is further constrained by the anatomy (via the network topology), yielding a signature of the brain dynamics. Instead of using directly EC as a biomarker, we have recently switched to a network-oriented analysis based on the estimated model, after fitting to data.
Results
In a recent application, we showed how our model-based approach uncovers differences between subjects with disorders of consciousness, from coma (UWS) to minimally conscious (MCI) and controls (awake) [2]. We find that the discrimination across patient types (and controls) can be quantitatively related to measuring whether the modeled stimulation response affects the whole network of brain regions. These results can further be interpreted in terms of over-segregation for UWS just after the stimulation, but more importantly a lack of integration in the sense of propagation of the response to the whole network late after the stimulation. In other words, we obtain personalized and interpretable biomarkers based on the brain dynamics.
Discussion
This framework can be used to quantify network complexity based on in-silico stimulation of a network model whose dynamics are estimated from ongoing data (i.e. without experimental stimulation). We will also discuss this approach with recent work based on statistical physics of out-of-equilibrium dynamic systems (related to time reversibility) that can also be interpreted in terms of network complexity [3].
Acknowledgements MG received support from the French government under the France 2030 investment plan, under the agreement Chaire de Professeur Junior (ANR-22-CPJ2-0020-01) and as part of the Initiative d’Excellence d’Aix-Marseille Université – A*MIDEX (AMX-22-CPJ-01).
