Fig. 3

Empirical inter-RSN disconnectivity (\(d_{inter}\))in stroke and healthy controls. \(d_\textrm{inter}\) reflects the fraction of inter-network disconnectivity - thus the loss of connections from a given RSN to the rest of the brain, normalized to healthy controls. Therefore, \(d_\textrm{inter}<1\) (decreased connectivity compared to controls average); \(d_\textrm{inter}=1\) (controls average); \(d_\textrm{inter}>1\) (increased connectivity compared to controls average). In total, 13 resting-state networks were considered. Each individual curve shows the rank of \(d_\textrm{inter}\) (from highest disconnectivity to lowest). Panels (a)-(c) show \(d_{inter}\) for individual participants in each group (patients at \(t_1\)/\(t_2\) and controls, respectively). Panel (d) shows the mean of the corresponding groups. Panels (e)-(h) show \(d_\textrm{inter}\) for artificial stroke newtorks according to⁴. In this particular example, we perform synthetic lesions in the default mode network (composed by 40 brain regions), visual network (38), auditory network (23), and sensory-motor mouth network (8), respectively. The fraction of nodes removed in all RSN is about \(80\%\) (see main text). In all panels (a)-(h), the average across control subjects is indicated in blue. Unlike the synthetic networks analyzed in⁴, empirical stroke networks present a more distributed disconnection profile. The individual profiles reveals that \(d_\textrm{inter}\) is almost uniformly distributed across RSN (see the smooth slope), i.e., the lesion affects almost uniformly all RSN in contradiction with the artificial strokes used in⁴. In this case, the profiles behave like a step-function, with an accentuated decrease of \(d_\textrm{inter}\) in the first RSN. Empirical inter-RSN disconnectivity patterns shows little or no change, as shown in the panel (d), where the lines for \(t_1\) and \(t_2\) are overlapping for all 13 RSNs (two tailed t-test: \(p>0.5\)). The asterisks in the plots represent significant two tailed t-tests (\(p<0.05\)) comparing real connectivity profiles in the control group with real lesions in stroke patients (a-d) and synthetic lesions (e-h).