Functional brain networks and cognitive deficits in Parkinson's disease

Abstract

Graph-theoretical analyses of functional networks obtained with resting-state functional magnetic resonance imaging (fMRI) have recently proven to be a useful approach for the study of the substrates underlying cognitive deficits in different diseases. We used this technique to investigate whether cognitive deficits in Parkinson's disease (PD) are associated with changes in global and local network measures. Thirty-six healthy controls (HC) and 66 PD patients matched for age, sex, and education were classified as having mild cognitive impairment (MCI) or not based on performance in the three mainly affected cognitive domains in PD: attention/executive, visuospatial/visuoperceptual (VS/VP), and declarative memory. Resting-state fMRI and graph theory analyses were used to evaluate network measures. We have found that patients with MCI had connectivity reductions predominantly affecting long-range connections as well as increased local interconnectedness manifested as higher measures of clustering, small-worldness, and modularity. The latter measures also tended to correlate negatively with cognitive performance in VS/VP and memory functions. Hub structure was also reorganized: normal hubs displayed reduced centrality and degree in MCI PD patients. Our study indicates that the topological properties of brain networks are changed in PD patients with cognitive deficits. Our findings provide novel data regarding the functional substrate of cognitive impairment in PD, which may prove to have value as a prognostic marker.

Keywords: Parkinson's disease; cognitive impairment; connectivity; fMRI; graph theory.

Figure 1

Comparisons of interregional connectivity strength between HC and PD subgroups. Top row: mean connectivity matrices according to group (Pearson's r coefficients, indicated in the color bar) for all pairs of ROIs. Bottom row: edges with significant post‐hoc group differences (P < 0.05) in z correlation values are marked in color. Bottom left: non‐MCI versus MCI patients; bottom middle: HC vs. MCI patients; bottom right: HC vs. non‐MCI patients. Color bars indicate post‐hoc Bonferroni test P values according to direction of differences. Anatomical regions, ordered in roughly anterior‐posterior sequence and grouped according to lobes or subcortical structures, are numbered in the vertical and horizontal axes according to Table 1. CS: corpus striatum; thal: thalamus.

Figure 2

Small‐world, modularity, and clustering coefficients (vertical axis) as a function of sparsity thresholds (horizontal axis) for HC and PD subgroups. * Indicates significant differences between HC and MCI patients (P < 0.05, post‐hoc Bonferroni test); § indicates significant differences between non‐MCI and MCI patients (P < 0.05, post‐hoc Bonferroni test). SW: small‐world coefficient. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Figure 3

Community structure and hub distribution according to group. Colors in nodes and links correspond to the modules indicated at the top of the figure. Black nodes in the HC group represent the thalami. Only intramodular edges are shown. Mean number of modules and SDs per group are shown at the right. Nodes classified as hubs in each group are listed, as well as the communities to which they belong in each group's mean network.

Figure 4

Changes in measures of node degree and BC in non‐MCI and MCI PD patients relative to HC as a function of HC's means. Left side: mean differences between PD‐non‐MCI and HC (A, C) and between PD‐MCI and HC (B, D) for node degree, K, (A, B) and BC (B, D) are plotted against the mean values in the HC group (horizontal axes) for all nodes. Nodes classified as hubs in the control group are numbered (see Table 1). Colors indicate the modules to which each node belongs in HC as indicated at the top of the figure. Right side: correlation between HC's mean values for K (above) and BC (below; horizontal axes) and the individual differences between subjects' values in the corresponding parameters and mean HC values (vertical axes). Mean r correlation values and SDs according to group are shown. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Figure 5

Changes in nodal parameters of segregation in non‐MCI and MCI PD patients relative to HC as a function of HC's means. Mean differences between PD‐non‐MCI and HC (A, C) and between PD‐MCI and HC (B, D) for node nodal clustering coefficients, C, (A, B) and local efficiency, eLoc, (B, D) are plotted against the mean values in the HC group (horizontal axes) for all nodes. Nodes classified as hubs in HC are numbered (see Table 1). Colors indicate the modules to which each node belongs in the HC group as indicated at the top of the figure. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]

Figure 6

Relationship between network parameters and composite scores for memory and VS/VP functions. A: Significant linear regression analysis results for global measures and VS/VP (top) and memory (bottom) z scores. β: standardized beta regression score. B: Regions where regional measures correlated significantly with A/E, memory, or VS/VP scores. Top row: negative correlations; bottom row: positive correlations. C: clustering coefficient; Mod: modularity; SW: small‐world coefficient; eLoc: local efficiency. [Color figure can be viewed in the online issue, which is available at http://wileyonlinelibrary.com.]