Poster E82, Monday, March 26, 2:30-4:30 pm, Exhibit Hall C
Multimodal structural predictors of naming therapy outcomes in persons with aphasia
Erin Meier1, Jeffrey Johnson1, Yue Pan1, Maria Dekhtyar1, Swathi Kiran1; 1Boston University
Cross-sectional studies (e.g., Sims et al., 2016; Basilakos et al., 2014) implicate the integrity of specific gray matter (GM) and white matter (WM) regions of interest (ROIs) in language recovery after stroke. The goal of this study was to determine the extent to which regional left hemisphere (LH) and right hemisphere (RH) integrity predicts naming therapy gains in persons with aphasia (PWA). Before therapy, 27 PWA underwent whole-brain DTI and T1-weighted scans. Bilateral ROIs implicated in naming (i.e., anterior cingulate; superior, middle and inferior frontal gyri; middle and inferior temporal gyri; supramarginal and angular gyri) (Indefrey & Levelt, 2004) were extracted from the Harvard-Oxford atlas. LH GM metrics (i.e., percent spared cortical tissue) and bilateral WM scalars (i.e., fractional anisotropy [FA], mean diffusivity [MD]) were obtained from cortical and subcortical masks in each ROI. To identify a cohesive set of predictors, LH and RH structural metrics were entered into two separate principal component analyses (PCAs). The LH and RH PCAs yielded six and three components, respectively. These nine components (i.e., LH_Temporal, LH_Parietal, LH_IFG, LH_DLPFC, LH_DMPFC, LH_ACC.spared, RH_FA, RH_MD, RH_Temporal_FA) were entered as predictors into a backward stepwise regression. The final model containing LH_Temporal, LH_IFG, LH_DMPFC, LH_DLPFC, and RH_MD explained 61% of the variance in treatment response (F(5,21)=6.49, p<.001). This multimodal model better predicted treatment gains than models containing GM metrics (p=0.003) or lesion volume (p=.024) alone. In sum, naming treatment success was most reliant on the combined GM and WM integrity of critical LH frontal and temporal regions.
Topic Area: METHODS: Neuroimaging