Table 1.
Demographic characteristics of the stroke patients included in our additional analysis.
Table 2.
Summary description of the different lesion methods used and their characteristics.
Table 3.
Summary description of Single-choice method and Hubs method for the selection of the N nodes to remove (see 2.5.1).
Figure 1.
Relationship between communicability, standard connectivity and distance measures.
A) Boxplot of the correlations of C and Cmw respectively within subjects (WS) intra-scan, WS inter-scan, and between subjects (BS). B) Scatter plot of Cmw and DistW matrices for the average network. C) Binary communicability assortativity matrix, i.e. Cm distribution among nodes with increasing Deg. D) Weighted communicability assortativity matrix, i.e. Cmw distribution between nodes with increasing Sw.
Figure 2.
Analysis of hubs and nodes with highest communicability.
A) maps of the nodes with highest Deg and/or Cm. B) maps of nodes with highest Sw and/or Cmw (normalized). The metric values for the nodes represented are at least one standard deviation (SD) over the average value. C) Density variation among the nodes with highest Deg, Cm, Sw, Cmw.
Figure 3.
Average efficiency decay (Eff, Effw) curves over subjects for the different target selection strategies.
Curves are reported for a total of N = 80 consecutives attacks.
Figure 4.
Local changes due to simulated lesions.
Top: number of significant local changes for the various metrics when hubs were targeted for binary lesions. The line types indicate the rate of deleted connections: solid lines R = 0.8, dashed lines R = 0.5 and dotted lines R = 0.2. Center: average number of significant local changes over 25 repetitions for the various metrics when random nodes were targeted for binary lesions. The line types indicate the site selection and rate: solid lines with cross markers for same nodes for all subjects and rate R = 0.8, dashed lines with cross markers for same nodes for all subjects and R = 0.5, dotted lines for with cross markers for same nodes for all subjects and R = 0.2 and solid lines with circle markers different nodes for each subject R = 0.8. Bottom: average number of significant local changes over 25 repetitions for the various metrics when edges were targeted for binary lesions. The line types indicate if the same edges (solid lines) or different edges (dashed lines) were selected for each subject.
Figure 5.
Relationship between the average local changes over subjects and the distance from the lesion site (after 5 attacks to hubs nodes).
Each dot in the figure represents a node in the average network. A positive change indicates a reduction in the metrics after the lesions. Scatter plots are reported for Sw (A), Cmw (B) and CBC (C) against weighted distance and for CBC against binary distance (D).
Table 4.
Correlation coefficients between local metric changes and (weighted and binary) distance from lesions are reported.
Figure 6.
Global and hemispheric network metrics of Deg, Cm, Sw and Cmw for healthy controls (HC) against stroke patients (SP).
Figure 7.
Local correlations between changes ((A) Deg, (B) Sw (C) Cmw (D) CBC) from the control group and distance from lesion for stroke patient L1 (largest lesion) and R1 (smallest lesion).
Crosses indicate local values for patient L1, while dots are associated to local values of patient R1. Lines indicate the least square lines associated to each relation.
Table 5.
Correlation coefficients of local differences from the control group with the Euclidean distance from the lesion are reported for each stroke patient and each network metric.