Computational modeling of resting-state activity demonstrates markers of normalcy in children with prenatal or perinatal stroke

Abstract

Children who sustain a prenatal or perinatal brain injury in the form of a stroke develop remarkably normal cognitive functions in certain areas, with a particular strength in language skills. A dominant explanation for this is that brain regions from the contralesional hemisphere "take over" their functions, whereas the damaged areas and other ipsilesional regions play much less of a role. However, it is difficult to tease apart whether changes in neural activity after early brain injury are due to damage caused by the lesion or by processes related to postinjury reorganization. We sought to differentiate between these two causes by investigating the functional connectivity (FC) of brain areas during the resting state in human children with early brain injury using a computational model. We simulated a large-scale network consisting of realistic models of local brain areas coupled through anatomical connectivity information of healthy and injured participants. We then compared the resulting simulated FC values of healthy and injured participants with the empirical ones. We found that the empirical connectivity values, especially of the damaged areas, correlated better with simulated values of a healthy brain than those of an injured brain. This result indicates that the structural damage caused by an early brain injury is unlikely to have an adverse and sustained impact on the functional connections, albeit during the resting state, of damaged areas. Therefore, these areas could continue to play a role in the development of near-normal function in certain domains such as language in these children.

Keywords: functional connectivity; network model; prenatal/perinatal stroke; resting-state fMRI; structural connectivity.

Figure 1.

Lesion masks drawn in native space. Participants indicated with an asterisk next to their participant number have lesions in the periventricular white matter, where it is not possible to delineate accurately the lesioned area from the ventricle. Participant 05's lesion is too small to be noticeable at this resolution, so its location is indicated with a red arrow.

Figure 2.

1000-ROIs parcellation. a, Brain-surface representation of the 1000-ROIs parcellation generated according to Cammoun et al. (2012). b, Histogram of the average ROIs' volume for the 10 healthy adults included in this study. The average volumes was 516.66 ± 187.41 mm³, corresponding to 10.16 ± 3.68 for a voxel size equal to 3.4 × 3.4 × 4.4 mm.

Figure 3.

Figure displays the correlations between FC values of each neonatal stroke participant with the corresponding values averaged across 27 AMC participants. a, Correlation between FC values of all (black) and only the damaged (gray) among the 68 brain areas with all other areas. b, Correlation between FC values of all (black) and only the damaged (gray) among the 998 ROIs with all other ROIs. None of the 998 ROIs in Participant 05 had a damage percentage greater than the 10% threshold that we applied in this study. Therefore, none of the 68 brain areas was considered damaged. Therefore, the correlations using these selected ROI/brain area pairs were not calculated for this participant.

Figure 4.

a, Blue curve displays the correlation between empirical FC values of all connections averaged across 27 AMC participants and the corresponding FC values simulated using the SC averaged across 10 healthy adults as a global coupling parameter, G, is varied. The red curve displays the correlation between empirical FC values from Participant 01 and FC values simulated using a lesioned SC matrix corresponding to the damage suffered by Participant 01. The value of G where this correlation is maximum is chosen as the optimal working point of the model. b, c, Scatter plots displaying the comparison between the empirical and the simulated FC values evaluated at the optimal working point of the model in each case for an average healthy participant (b) and for Participant 01 (c).

Figure 5.

a, Correlation between the empirical FC values of each of the 68 brain areas with all other areas for every participant with the corresponding simulated FC values using a healthy SC matrix (R-H) and using the corresponding lesioned/damaged SC matrix (R-L). Bottom, Mean (±SEM) correlations averaged across all participants (b) and averaged across only the most severely affected participants (Participants 01, 06, 07, 08, 09, and 10) (c). The differences in the mean correlations, whether across all participants (p = 0.12, paired t test) or across strongly damaged participants (p = 0.17, paired t test), were not statistically significant.

Figure 6.

a, Correlation between the empirical FC values of only the damaged among the 68 brain areas with all other areas for every participant with the corresponding FC values simulated using a healthy SC matrix (R-H) and using the corresponding lesioned/damaged SC matrix (R-L). Bottom, Mean (±SEM) correlations averaged across all participants (b) and averaged across only the most affected participants (Participants 01, 06, 07, 08, 09, and 10) (c). None of the 998 ROIs in Participant 05 had a damage percentage greater than the 10% threshold that we applied in this study. Therefore, none of the 68 brain areas was considered damaged, and therefore the correlations using these selected brain area pairs were not calculated for this participant. Although differences between mean correlations across all participants were not statistically significant (p = 0.09, paired t test), they were statistically significant when means are averaged across only the most affected participants (p = 0.01, paired t test).

Figure 7.

a, Correlation between the empirical FC values of only the contralesional among the 68 brain areas with all the other brain areas, for every participant, with the corresponding FC values simulated using a healthy SC matrix (R-H) and using the corresponding lesioned/damaged SC matrix (R-L). Bottom, Mean (±SEM) correlations averaged across all participants (b) and averaged across only the most affected participants (Participants 01, 06, 07, 08, 09, and 10) (c). None of the 998 ROIs in Participant 05 had a damage percentage greater than the 10% threshold that we applied in this study. Therefore, none of the 68 brain areas was considered damaged, and therefore the correlations using these selected brain area pairs were not calculated for this participant. The differences in the mean correlations, whether across all participants (p = 0.9, paired t test) or across the most affected participants (p = 0.17, paired t test), were not statistically significant.

Figure 8.

a, Correlation between the empirical FC values of only the damaged among the 998 ROIs with all other ROIs for every participant with the corresponding FC values simulated using a healthy SC matrix (R-H) and using the corresponding lesioned/damaged SC matrix (R-L). Bottom, Mean (±SEM) correlations averaged across all participants (b) and averaged across only the most affected participants (Participants 01, 06, 07, 08, 09, and 10) (c). None of the ROIs in Participant 05 had a damage percentage greater than the 10% threshold that we applied in this study, and therefore the correlations using these selected ROI pairs were not calculated for this participant. The differences in the mean correlations, whether across all participants (p = 0.02, paired t test) or across the most affected participants (p = 0.007, paired t test), were strongly statistically significant.

Figure 9.

Results from the second approach of lesion modeling: mean (±SEM) correlations averaged across all participants (a–c) and averaged across only the most affected participants (Participants 01, 06, 07, 08, 09, and 10) (d–f). Correlations between the empirical FC values and the corresponding FC values simulated using a healthy SC matrix (R-H) and using the corresponding lesioned/damaged SC matrix (R-L) were obtained using the FC values of all pairs of brain areas (a, d), only the damaged areas with all other areas (b, e), and only the contralesional areas with all other areas (c, f). The differences in the mean correlations, obtained using FC values of damaged areas with all other areas (b, e), whether across all participants (p = 0.02, paired t test) or across the most affected participants (p = 0.02, paired t test), were statistically significant.

Figure 10.

Results from the third approach of lesion modeling: mean (±SEM) correlations averaged across all participants (a–c) and averaged across only the most affected participants (Participants 01, 06, 07, 08, 09, and 10) (d–f). Correlations between the empirical FC values and the corresponding FC values simulated for the healthy case (R-H) and for the corresponding lesioned case (R-L) were obtained using the FC values of all pairs of brain areas (a, d), only the damaged areas with all other areas (b, e), and only the contralesional areas with all other areas (c, f). The differences in the mean correlations, obtained using FC values of damaged areas with all other areas (b, e), whether across all participants (p = 3E-3, paired t test) or across the most affected participants (p = 4E-4, paired t test), were strongly statistically significant. This very high difference also contributed to statistically significant difference between mean correlations, calculated using all FC values (a, d) averaged across all participants (p = 0.02, paired t test), as well as across the most affected participants (p = 3E-3, paired t test).