Introduction

Numerous studies have demonstrated that the relation between brain changes and functional impairment and recovery should be studied from the perspective of brain network ¹. Considering that there exists a group of brain regions with large number of connections, known as rich-club organization, that may govern brain functions ^2–4, we hypothesize that such rich-club organization may be sensitive to brain motor recovery after stroke. In this study, we therefore aimed to investigate rich-club organizational changes over the course of motor recovery for patients with first-time acute ischemic stroke in the motor system.

Methods

MRI experiments (3.0T Achieva TX, Philips, Best, Netherland) were performed and motor functions (upper-extremity Fugl-Meyer motor scale, UE-FM, and Barthel index, BI) were measured from n = 17 stroke patients (age: 44 – 84; infarct side: 33% left) at within 1 week, and 1, 3 and 6 months after stroke (n = 13, 16, 13, 9, respectively). DTI was performed using SS-SE-EPI with TR/TE = 4000/81 ms, b-value = 1000 s/mm2, 32 diffusion directions, and 3 mm isotropic resolution. The pipeline for the estimation of structural connectome (fiber count as weights) from DTI data is the same as that previously published ⁵. Group-averaged connectome was subsequently obtained for each time point (connections that existed in no more than 75% of all subjects were zeroed).
To determine the existence of rich-club organization, rich-club coefficient of a subnetwork with a given threshold of nodal degree as compared to that of a random network with the same degree distribution and weights (permutation testing with n = 10,000) was computed ^6,7. All brain regions were divided into rich-club and peripheral nodes. The local efficiency, local clustering coefficient and betweenness centrality of these two node types were measured. All brain connections were divided into rich-club (between rich-club nodes), feeder (between rich-club and peripheral nodes), and local (between peripheral nodes) connections ⁸. Network density (i.e. fiber count), network cost (product between fiber count and length), and communication cost (based on the topological distance between any brain region pair) were estimated ⁸. The cost/density ratio and communication cost/density ratio of three connection types were estimated to evaluate cost and capacity with imbalanced rich-club nodes.
Two-way ANOVA (Between subject factor: rich-club and peripheral nodes; or rich-club, feeder, and local connections. Time factor: time after stroke) with post-hoc analysis were performed. One-way repeated measures ANOVA was also performed to investigate the effect of time on motor recovery.

Results and Discussion

One-way repeated measures ANOVA showed that the BI (F = 20.20, p < 0.01) and UE-FM (F = 6.68, p = 0.0295) increased with the time after stroke. The BI measured at 3 and 6 months were higher than baseline at less than a week after stroke (p < 0.01 and < 0.001, respectively), whilst UE-FM were not (p = 0.1176 and 0.0936). Rich-club organization was found from the group-averaged structural brain network at all time points, except within a week after acute stroke (Figure 1). These rich-club nodes are tabulated in Table 1. These results showed that rich-club organization remerged one month after stroke. Higher order brain regions, those with larger nodal degree, contributed to part of the rich-club organization only at 6 months post stroke, at which point patients already functionally recovered.
There was no interaction effect between time and node types on the local efficiency (F = 0.750, p = 0.473), local clustering coefficient (F = 0.782, p = 0.458) and betweenness centrality (F = 2.938, p = 0.055). There was a main effect of node type on local clustering and rich-club nodes had lower local clustering coefficient (F = 15.221, p < 0.001) compared with peripheral nodes. There was a main effect of time on betweenness centrality (F = 23.532, p < 0.001) and it decreased in the order - 1, 3 and 6 months after stroke. There was an interaction effect between time and connection types on the cost/density ratio (F = 2.801, p = 0.025). As shown in Figure 2, there was a simple main effect of time on the cost/density ratio of rich-club connections (F = 4.174, p = 0.015), whereby the cost/density ratio at 6 months post stroke was higher than 3 months (p = 0.012). For each time point, the cost/density ratio decreased in the order - rich-club (high), feeder (middle) and local (low) connections (p < 0.01).
When examining all of the shortest communication paths (Figure 3), cost of rich-club connections accounted for an increasing percentage of total communication cost (from 39.85%, 45.87% to 48.95%), while feeder connections accounted for a decreasing percentage (from 53.81%, 50.19% to 42.85%). Both of rich-club (from 1.09, 1.20 to 1.00) and feeder connections (from 1.26, 1.13 to 1.20) showed a fluctuate capacity over the course of stroke recovery. Combining cost/density ratio, feeder connections showed a phenomenon of middle cost and high capacity while rich-club connections showed high cost and middle capacity when patients functionally recovered. These results are different from the high-cost and high-capacity of rich-club organizations of healthy controls as previously reported ⁸.

Conclusion

Our results suggest that there could be a potential relation between reemergence of rich-club organization and brain motor recovery.

Acknowledgements

No acknowledgement found.