Functional Hyperconnectivity during a Stories Listening Task in Magnetoencephalography Is Associated with Language Gains for Children Born Extremely Preterm

Abstract

Extreme prematurity (EPT, <28 weeks gestation) is associated with language problems. We previously reported hyperconnectivity in EPT children versus term children (TC) using magnetoencephalography (MEG). Here, we aim to ascertain whether functional hyperconnectivity is a marker of language resiliency for EPT children, validating our earlier work with a distinct sample of contemporary well-performing EPT and preterm children with history of language delay (EPT-HLD). A total of 58 children (17 EPT, 9 EPT-HLD, and 32 TC) participated in stories listening during MEG and functional magnetic resonance imaging (fMRI) at 4-6 years. We compared connectivity in EPT and EPT-HLD, investigating relationships with language over time. We measured fMRI activation during stories listening and parcellated the activation map to obtain "nodes" for MEG connectivity analysis. There were no significant group differences in age, sex, race, ethnicity, parental education, income, language scores, or language representation on fMRI. MEG functional connectivity (weighted phase lag index) was significantly different between groups. Preterm children had increased connectivity, replicating our earlier work. EPT and EPT-HLD had hyperconnectivity versus TC at 24-26 Hz, with EPT-HLD exhibiting greatest connectivity. Network strength correlated with change in standardized scores from 2 years to 4-6 years of age, suggesting hyperconnectivity is a marker of advancing language development.

Keywords: connectivity; development; language; magnetic resonance imaging; magnetoencephalography; prematurity.

Figure 1

Joint functional MRI activation map and extraction of virtual sensors. (A) ANOVA was performed in SPM12 to assess differences in language representation (clusters of voxel-wise activation). There were no statistically significant differences between groups. To objectively identify the language network in our cohort, we then computed the joint activation map across groups (as there were no group differences in representation detected). Sixteen axial slices from the fMRI joint activation map (EPT-HLD + EPT + TC) are shown with typical bilateral activation in response to language stimuli (auditorily presented passive stories listening) versus noise condition (p < 0.001, k = 8). “L” denotes the left side of the brain in all images. (B) The joint activation map from fMRI was parcellated using a 200-unit random parcellation scheme. Centroids of parcels with significant activation (greater than 10% active voxels) served as “nodes” for subsequent connectivity analyses, shown in blue, that was performed on MEG data obtained during the same stories listening task.

Figure 2

MEG functional connectivity indexed by weighted phase lag index for all preterm children and all term children. Weighted phase lag index (wPLI) extracted from timeseries at virtual sensors shown in Figure 1. All extremely preterm children (EPT-HLD + EPT, n = 26) are shown in pink, and term children (TC, n = 32) are shown in black. Statistically significant differences in connectivity between groups include increased functional connectivity for preterm participants at 14.5–15.5 Hz, 24.5–26.5 Hz, and 30–31.5 Hz. Images were generated in MATLAB (2020).

Figure 3

MEG functional connectivity indexed by weighted phase lag index for three groups: Extremely preterm with history of formally diagnosed language delay (EPT-HLD), extremely preterm without language delay (EPT), and term children (TC). Weighted phase lag index (wPLI) extracted from timeseries at virtual sensors shown in Figure 1. Functional connectivity between all three groups (TC in black, EPT-HLD in red, and EPT in blue) was significantly different as assessed by a one-way ANOVA (p < 0.05). EPT and EPT-HLD showed functional hyperconnectivity versus TC at 24–26 Hz with EPLI having highest functional connectivity. Images were generated in MATLAB (2020).

Figure 4

Significant subnetworks supporting hyperconnectivity in preterm children. Investigation of significant between groups differences (all preterm > TC) demonstrated a more bilateral network with the bulk of connections traversing the hemispheres. When significant differences were investigated between the EPT-HLD group and the EPT and TC groups in these post hoc comparisons, the significant connections appeared to be more concentrated in the left perisylvian region. “R” denotes the right side of the brain in all images. Images were generated from the adjacency matrices exported from NBS using the CONN toolbox running in MATLAB (2019b). (A) All extremely preterm children (EPT-HLD + EPT) collectively demonstrate significant subnetwork supporting hyperconnectivity versus term children (TC). Network “edges” showing significantly increased functional connectivity in all preterm participants versus TC between 24 and 26 Hz during stories listening (observed at various initial thresholds ranging from t =0.1 to 3.8, median t value of 1.95 shown, 5000 permutations, p < 0.05, corrected for multiple comparisons). (B) Extremely preterm children with a history of language delay (EPT-HLD) exhibit significant subnetwork supporting hyperconnectivity versus term children (TC). Network “edges” showing significantly increased functional connectivity in EPT-HLD versus TC between 24 and 26 Hz during stories listening (observed at various initial thresholds ranging from t =1 to 5, median t value of 3 shown, 5000 iterations, p < 0.05, corrected for multiple comparisons). (C) Extremely preterm children with a history of language delay (EPT-HLD) exhibit significant subnetwork supporting hyperconnectivity versus extremely preterm peer without language delay (EPT). Network “edges” showing significantly increased functional connectivity in EPT-HLD versus EPT between 24 and 26 Hz during stories listening (observed at various initial thresholds ranging from t =2.2 to 3.5, median t value of 2.9 shown, 5000 iterations, p < 0.05, corrected for multiple comparisons).