Structural and effective connectivity of lexical-semantic and naming networks in patients with chronic aphasia

Given the difficulty in predicting outcomes in persons with stroke-induced aphasia (PWA), neuroimaging-based biomarkers of recovery could provide invaluable predictive power to stroke models. However, the neural patterns that constitute beneficial neural organization of language in PWA remain debated. Thus, in this work, we propose a novel network theory of aphasia recovery and test our overarching hypothesis, i.e., that task-specific language processing in PWA requires the dynamic engagement of intact tissue within a bilateral network of anatomically-segregated but functionally and structurally connected language-specific and domain-general brain regions. We first present two studies in which we examined left frontotemporal connectivity during different language tasks (i.e., picture naming and semantic feature verification). Results suggest that PWA heavily rely on left middle frontal gyrus (LMFG)-driven connectivity for tasks requiring lexical-semantic processing and semantic control whereas controls prefer models with input to either LMFG or left inferior frontal gyrus (LIFG). Both studies also revealed several significant associations between spared tissue, connectivity and language skills in PWA. In the third study, we examined bilateral frontotemporoparietal connectivity and tested a lesion- and connectivity-based hierarchical model of chronic aphasia recovery. Between-group comparisons showed controls exhibited stronger left intra-hemispheric task-modulated connectivity than did PWA. Connectivity and language deficit patterns most closely matched predictions for patients with primarily anterior damage whereas connectivity results for patients with other lesion types were best explained by the nature of the semantic task. In the last study, we investigated the utility of lesion classification based on gray matter (GM) only versus combined GM plus white matter (WM) metrics. Results suggest GM only classification was sufficient for characterizing aphasia and anomia severity but the GM+WM classification better predicted naming treatment outcomes. We also found that fractional anisotropy of left WM association tracts predicted baseline naming and treatment outcomes independent of total lesion volume. Finally, results of a preliminary multimodal prediction analysis suggest that combined structural and functional metrics reflecting the integrity of regions and connections comprise optimal predictive models of behavior in PWA. To conclude this dissertation, we discuss how multimodal network models of aphasia recovery can guide future investigations.2020-10-23T00:00:00