Normalization of network connectivity in hemispatial neglect recovery

Abstract

Objective: We recently reported that spatial and nonspatial attention deficits in stroke patients with hemispatial neglect are correlated at 2 weeks postonset with widespread alterations of interhemispheric and intrahemispheric functional connectivity (FC) measured with resting-state functional magnetic resonance imaging across multiple brain networks. The mechanisms underlying neglect recovery are largely unknown. In this study, we test the hypothesis that recovery of hemispatial neglect correlates with a return of network connectivity toward a normal pattern, herein defined as "network normalization."

Methods: We measured attention deficits with a neuropsychological battery and FC in a large cohort of stroke patients at, on average, 2 weeks (n = 99), 3 months (n = 77), and 12 months (n = 64) postonset. The relationship between behavioral improvement and changes in FC was analyzed both in terms of a priori regions and networks known to be abnormal subacutely and in a data-driven manner.

Results: Attention deficit recovery was mostly complete by 3 months and was significantly correlated with a normalization of abnormal FC across many networks. Improvement of attention deficits, independent of initial severity, was correlated with improvements of previously depressed interhemispheric FC across attention, sensory, and motor networks, and a restoration of the normal anticorrelation between dorsal attention/motor regions and default-mode/frontoparietal regions, particularly in the damaged hemisphere.

Interpretation: These results demonstrate that abnormal network connectivity in hemispatial neglect is behaviorally relevant. A return toward normal network interactions, and presumably optimal information processing, is therefore a systems-level mechanism that is associated with improvements of attention over time after focal injury. Ann Neurol 2016;80:127-141.

Figure 1

Topography of stroke. Lesion overlap image in atlas space for all patients (A), patients without and with neglect (B) and all patients, with lesions flipped to the right side (C). The color scale indicates the number of subjects with lesions at each voxel.

Figure 2

Neglect and recovery. The distribution of neglect scores in controls and in patients sub-acutely and 3 months and 1 year post stroke (A). Visual attention deficit (VAD) scores over time for patients with and without sub-acute deficits and controls (B). Correlation of VAD scores with scores in other behavioral domains at the sub-acute timepoint (C). Correlation of the improvement (3 months minus sub-acute) of VAD scores with the improvement of scores in other behavioral domains (D).

Figure 3

Right medial IPS functional connectivity (FC) with the rest of the brain in relation to behavior. FC by behavior maps: sub-acutely post stroke (B), recovery (E) and recovery with sub-acute behavior regressed out (H). Red/yellow indicates a positive correlation and green/blue a negative correlation. Scatter plot of FC versus behavior for an inter-hemispheric region (A, D, G) and intra-hemispheric region (C, F, I) highlights the correlations. Each dot is a patient and blue indicates patients with right hemisphere lesions, red with left hemisphere lesions.

Figure 4

DAN and DAN-DMN functional connectivity (FC). Inter-hemispheric DAN (A), left hemisphere DAN to DMN (B) and right hemisphere DAN to DMN (C) FC over time for the deficit (red) and no deficit (blue) patients and controls (green). The brain figures show a subset of the connections that were averaged in the graphs (blue dots represent DAN ROIs and red dots DMN ROIs).

Figure 5

Inter- and intra-hemispheric functional connectivity (FC). Homotopic FC (A) for each network and intra-hemispheric FC for the DMN versus other networks in the left (B) and right (C) hemisphere over time for the deficit (dashed line) and no deficit (solid line) patients and controls (dotted line).

Figure 6

Behaviorally relevant functional connectivity (FC) recovery of the left hemisphere. The loadings for each of the 10 networks of the first component of the principal component analysis of voxel-wise FC and behavior change maps of the left hemisphere (A). The voxelwise FC change maps of the top 10% ROIs (depicted in B) are averaged and correlated with the VAD change. The inter-hemispheric FC to VAD change is shown on the left hemisphere (C) and the within hemisphere FC to VAD change is shown on the left hemisphere (D). On the lateral views of the brain red/yellow indicates a positive FC-VAD correlation and green/blue a negative correlation (Z-statistic of Pearson r, thresholded at |Z|>2.25, P<0.05, Monte Carlo corrected). The scatterplots depict the correlation between the average FC change of the top 10% seeds and the black circle and behavioral improvement (C: Supramarginal gyrus, Talairach coordinates: +45 −38 +34; D: PFC, −32 +46 +23). Each dot is a patient and blue indicates patients with right hemisphere lesions, red with left hemisphere lesions. Networks: Networks: VFN = visual foveal representation; VPN = visual peripheral representation;DAN = dorsal attention; MN = motor; AN = auditory; VAN = ventral attention; CON = cingulo-opercular; LN = language;FPN = frontoparietal; DMN = default mode. Labels: FEF, frontal eye fields;IPS, intraparietal sulcus; TPJ, temporal parietal junction; pIns, posterior insula; aIns, anterior insula; MT+, middle temporal complex; PFC, prefrontal cortex; dPoCe, dorsal post central gyrus.

Figure 7

R mIPS correlations with recovery of other deficits in patients with right hemisphere lesions. Whole brain functional connectivity improvement correlations of the right medial IPS with behavioral recovery of A. Attention (n=37), B. Language (n=34) and C. Motor (n=34) recovery. In each of these correlations the sub-acute behavioral deficits are regressed out. In D attention recovery is depicted, while regressing out the left limb motor recovery (n=34). E shows the inter-hemispheric dorsal attention network (DAN) and intra-hemispheric DAN to default mode network (DMN) connectivity changes correlated with attention, language and motor improvement as well as attention with language and motor regressed out.

Figure 8

Lesion type, Lag and Lesion size. The scatterplots show the correlation between behavior (VAD) change and FC change between sub-acute and 3 months post stroke for the average of the homotopic DAN nodes (A), the correlations of VAD change and FC change separated out by lesion type: subcortical, cortical and cortical-subcortical (B), the correlation between VAD change and FC change when lesion size is regressed out (C) and the correlation between VAD change and FC change with lag change regressed out (D). Each dot is a patient and grey triangles indicate patients with right hemisphere lesions, black circles patients with left hemisphere lesions.