A data-driven definition of metastable brain states and trajectories characterizes group differences in various brain processes

Motivation: Analyzing brain activity as trajectories in a continuous neural dynamical space or as transitions in a discrete neural state-space is not settled on. Goal(s): To propose an approach from neural activity that identify discrete brain configurations. Approach: Combining BOLD extraction by ICA, reconstruction of neuronal activity and uniform manifold approximation and projection. We illustrated the results on three groups. Results: We identified distinct brain configurations highly shared among participants. Specific state sequences showed strong group differences. We suggest that the use of Nact, instead of BOLD, contributed to the good segregation of the different brain configurations by UMAP and allowed individual trajectories observation. Impact: Reproducibility of these highly segregated brain configurations between subjects supports the adequacy of models describing brain dynamics as transitions in a multistable state-space. Our method comes with an appealing temporal resolution to study pathological or/and individual brain activity trajectories.