Simulated Attack Reveals How Lesions Affect Network Properties in Poststroke Aphasia

Abstract

Aphasia is a prevalent cognitive syndrome caused by stroke. The rarity of premorbid imaging and heterogeneity of lesion obscures the links between the local effects of the lesion, global anatomic network organization, and aphasia symptoms. We applied a simulated attack approach in humans to examine the effects of 39 stroke lesions (16 females) on anatomic network topology by simulating their effects in a control sample of 36 healthy (15 females) brain networks. We focused on measures of global network organization thought to support overall brain function and resilience in the whole brain and within the left hemisphere. After removing lesion volume from the network topology measures and behavioral scores [the Western Aphasia Battery Aphasia Quotient (WAB-AQ), four behavioral factor scores obtained from a neuropsychological battery, and a factor sum], we compared the behavioral variance accounted for by simulated poststroke connectomes to that observed in the randomly permuted data. Global measures of anatomic network topology in the whole brain and left hemisphere accounted for 10% variance or more of the WAB-AQ and the lexical factor score beyond lesion volume and null permutations. Streamline networks provided more reliable point estimates than FA networks. Edge weights and network efficiency were weighted most highly in predicting the WAB-AQ for FA networks. Overall, our results suggest that global network measures provide modest statistical value beyond lesion volume when predicting overall aphasia severity, but less value in predicting specific behaviors. Variability in estimates could be induced by premorbid ability, deafferentation and diaschisis, and neuroplasticity following stroke.SIGNIFICANCE STATEMENT Poststroke, the remaining neuroanatomy maintains cognition and supports recovery. However, studies often use small, cross-sectional samples that cannot fully model the interactions between lesions and other variables that affect networks in stroke. Alternate methods are required to account for these effects. "Simulated attack" models are computational approaches that apply virtual damage to the brain and measure their putative consequences. Using a simulated attack model, we estimated how simulated damage to anatomic networks could account for language performance. Overall, our results reveal that global network measures can provide modest statistical value predicting overall aphasia severity, but less value in predicting specific behaviors. These findings suggest that more theoretically precise network models could be necessary to robustly predict individual outcomes in aphasia.

Keywords: WAB; aphasia; network; neuroimaging; simulated attack; stroke.

Figure 1.

Distribution of lesioned parcels across subjects with stroke. Lesions mapped prominently to parcels in left perisylvian regions with decreasing frequency in the superior, inferior, anterior, and posterior directions. The degree of red is proportional to the number of subjects with lesions at that location. N = the number of subjects with a lesion at that parcel label.

Figure 2.

Schematic of stroke imputation and diffusion tractography. A, Processing scheme for healthy subjects. Diffusion tractography was computed in subjects' native space, and the Lausanne multiscale parcellation was fit to subjects' anatomic T1 images. Connectomes were defined based on the FA or streamline counts of the edges connecting each region pair and advanced to analyses. B, The processing scheme for stroke subjects was the same as the healthy subjects with an additional preprocessing step. Specifically, the anatomic T1 image was imputed using the stroke subject's right hemisphere and healthy subjects' data to estimate the prelesion T1 anatomic image. The parcellation was computed on this imputed anatomic image to guide connectome extraction through the same regions as the controls.

Figure 3.

Schematic of network lesion masking. A, Top, Each element A_i j from each subject with aphasia (Stroke_i) was compared with (bottom) the observed values in all control subjects (Control₁ to Control_n). B, The elements with FA or streamlines 2, 3, or 4 standard deviations (SDs) less than controls were labeled as lesioned edges. C, Then, the lesion mask was applied to the stroke subject and all control subjects, and the resulting networks were advanced to connectomic analyses.

Figure 4.

Schematic of network measures. Moving Left to Right from the Top, We began with the (1) sum of edge weights in each network as an overall metric capturing the density of the networks, including any edges lost because of stroke. Four other measures were of interest. (2) Modularity: measures the extent to which nodes in the network are grouped into modules (sometimes, “communities”) as a function of highly-connected nodes. (3) Global efficiency: one long path is represented by the set of consecutive edges highlighted in green. (4) Transitivity: one possible triplet's edges are represented in green. (5) SWP: involves a high degree of local clustering (represented by the set of nodes connected by purple edges) and short path lengths (e.g., higher weights along the green path represents a shorter network path between the prefrontal and occipital nodes).

Figure 5.

The effects of stroke on network measures in observed and simulated attack connectomes. The leftmost column of each plot facet shows the network statistic observed in controls, followed by that observed in strokes, then the simulated attacks. Network measures are presented in their raw (untransformed) values before inclusion in network-behavior analyses. Asterisks indicate a significant Welch's two-sample t test between the control and stroke network measures at p < 0.001 (a stringent threshold after Bonferroni correction for 40 total tests in FA and streamline data). The top and bottom edges of the boxes represent the 25th and 75th percentiles, respectively. SWP = small world propensity.

Figure 6.

Network measures and behavioral variance in FA and streamline networks. Each plot facet illustrates histograms of the simulated null, histograms of the simulated attack distributions, and the observed R² with solid vertical lines. Asterisks indicate significant post hoc Welch's one-tailed t tests assuming unequal variances comparing the R² values in the simulated attacks to the null distribution at p < 0.001. Daggers indicate cases where the observed R² value was outside the range obtained in the simulated attack models. See Extended Data Figures 6-1, 6-2, 6-3, and 6-4 for specific effects of behavior on variance explained in the whole brain and left hemisphere.

Figure 7.

Simulated attack estimated beta weights for each network measure in FA and streamline networks. Each plot represents the range of betas obtained from the simulated attack models for the network measure. The top and bottom edges of the boxes represent the 25th and 75th percentiles, respectively. W. = edge weights; Mod. = modularity; Eff. = efficiency; Trans. = transitivity; SWP = small world propensity. See Extended Data Figure 7-1 for observed and simulated model betas.