CHARM captures the distinct network interactions between different brain states (IMAGE)
Caption
(a) The two matrices of the shifted functional connectivity between seven manifold networks reflect different interactions in wakefulness (top) and deep sleep (bottom). Importantly, there are significant differences (white squares) between the two matrices (p<0.05). (b) These time asymmetries are illustrated (for the top 20%) of arrows between the seven manifold networks for awake vs sleep, where the grey arrows correspond to negative and brown to positive interactions, and their thickness to the value of these interactions. As can be seen the flow of interaction is very different for wakefulness compared to deep sleep. Machine learning was able to separate these interactions with an accuracy of 84%. (c) The spatiotemporal characteristics in the CHARM latent space can be captured by the Edge metastability which is significantly different between awake and sleep (upper panel, p<0.05). Similarly, the reconfiguration of the hierarchical information flow is captured by the measure of network directedness of the shifted functional connectivity of the manifold network, which is also significantly different between awake and sleep (lower panel, p<0.01).
Credit
Deco, G.; Sanz Perl, Y.; Kringelbach M.L
Usage Restrictions
Credit must be given to the creator.
License
CC BY