Find in Library

Introduction. There is recent evidence that suggests that resting-state abnormalities in chronic stroke likely influence recovery patterns. In this study, we examined resting state networks in 11 patients with post-stroke aphasia (PWA) in comparison to a database of typical resting state-connectivity patterns to identify the relationship between resting state patterns and post-stroke language impairment and recovery. Methods. 11 PWA with left hemisphere strokes and at least four months post onset participated 3D anatomical images were acquired at TE/TR/TI = 3.04/2100/785 ms, flip angle = 13. The resting state fMRI (rs-fMRI) series contained 140 images with TR/TE/#slices/voxel size/slice thickness=3s/30ms/48/3.3125mmx3.3125mmx3.3125mm/3.3125mm with axial slice acquisition. A group of control subjects over the age of 50 were selected from the 1000 Functional Connectomes database (Biswal et al., 2011; http://www.nitrc.org/projects/fcon_1000). The rs-fMRI image series were pre-processed using standard procedures in SPM8. Reduction of noise of non-neural origin was performed using the CompCor method (Behzadi et al., 2007), removing signals related to the leading principal components of voxels from the white matter and CSF. The residual signals were bandpass filtered between 0.01 and 0.08 Hz in order to isolate the low frequency fluctuations. Voxels were masked to select those that are likely to be gray matter (GM), and non GM voxels were discarded from further analysis. Because the brains of PWA are likely to undergo post-stroke functional reorganization, we used a data-driven method to cluster voxels that (i) are spatially contiguous, and (ii) have time series that are correlated above a threshold. The resulting clusters served as nodes in a graph theoretic analysis, with weighted edges determined based on: i) the Pearson correlations between the mean cluster time series, and (ii) Partial correlations determined using a regularized version of the inverse covariance matrix (e.g., Friedman et al., 2008). Results. To test the hypothesis that there will be shifts in “network hubs” due to post-stroke recovery, we calculated the weighted node degree (sum of all edge weights impinging on a node / cluster) and betweenness centrality (the fraction of the shortest paths between all nodes in the graph that pass through that node) for each cluster, in both PWA and controls. Initial results show an increase in the average number of shortest paths that traverse each cluster in the graph in PWA relative to controls. Figure 1a and 1b show two examples of patient node degree maps, indicating shifts in network hubs (i.e., nodes with particularly high degree and/or betweenness centrality). To test the hypothesis that functional connectivity networks in PWA show decreased efficiency relative to controls, we calculated the global graph efficiency, defined as the inverse of the harmonic mean of the shortest path lengths between all nodes, for each subject. Initial results show a general pattern of reduced global efficiency in PWA relative to controls. Conclusions. Relative to controls, these results indicate inefficiencies in the post-stroke resting-state network, with greater shifts in network hubs in PWA dependent on the site and size of lesion. Such graph analytic results may prove informative in advancing individual-specific therapies.