You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

Poster B34, Wednesday, November 8, 3:00 – 4:15 pm, Harborview and Loch Raven Ballrooms

Left hemisphere frontotemporal effective connectivity during semantic feature judgments: Differences between patients with aphasia and healthy controls

Erin Meier1, Swathi Kiran1;1Boston University, Sargent College of Health and Rehabilitation Sciences

Making semantic judgments requires integrated functioning of anatomically-distributed regions implicated in low-level conceptual processing (e.g., left anterior/mid middle temporal gyrus [LMTG]), semantic control (e.g., left inferior frontal gyrus [LIFG]), and domain-general executive processes (e.g., left middle frontal gyrus [LMFG]) (e.g., Binder et al., 2009; Noonan et al., 2013). Left hemisphere (LH) stroke forces reorganization of this system. Compared to controls, persons with aphasia (PWA) may rely more heavily on domain-general regions due to difficulty with language and/or damage to canonical language cortex (Brownsett et al., 2014). However, little is known about the dynamic coupling between semantic network nodes in PWA. Therefore, we employed fMRI and dynamic causal modeling (DCM; Friston et al., 2003) to investigate differences between PWA and controls in effective connectivity of a three-node LH semantic network. Sixteen PWA (10M, mean age=64.9 years) and 17 age-matched controls (10M, mean age=60.4 years) completed an event-related fMRI task during which participants decided whether written features applied to real (experimental) or scrambled (control) pictures. MR and connectivity methods similar to Meier, Kapse, & Kiran (2016) were applied. In short, a DCM model space was created according to exogenous task-based perturbation to LIFG, LMFG or LMTG with all combinations of uni- or bidirectional inter-regional connections specified across 72 models. Models were partitioned into separate families (i.e., #1: LIFG input, #2: LMFG input, #3: LMTG input). Random effects family-wise Bayesian Model Selection (fw-BMS) determined which model family best fit individual and group data (Penny et al., 2010). Bayesian Model Averaging across each family generated parameter values indicating the strength/direction of task-based effects on regions and connections. Two-way (group x family) ANOVA and MANOVA were performed to determine group differences in regions and connections, respectively. Patients exhibited significantly poorer task accuracy than controls (t(17)=-3.32, p=.004). Group-level fw-BMS indicated models with task-induced perturbation to LIFG (family #1) best fit control data (xp=.881). Family #3: LMTG input best fit PWA group data (xp=.499), but only six PWA showed a preference for family #3 per single-subject fw-BMS. Patient model fit was not related to the amount of spared tissue in any region (r=-.025-.351, p=.199-.930). Regarding parameters, patients exhibited significantly weaker task-induced perturbation to all regions than controls (F(1,93)=16.02, p<.001). However, connection values were significantly more positive (i.e., excitatory) for PWA (F(6,88)=6.13, p<.001) for all connections except LIFG->LMFG and LMTG->LMFG. Overall, these findings illustrate that LH stroke alters the functional topography of the semantic network in PWA. Despite heterogeneity in model fit, patients relied more on lower-level processing by LMTG to drive semantic decisions whereas controls relied on LIFG, consistent with semantic control requirements for successful task completion (Thompson-Schill et al., 1997). PWA also exhibited weaker local task-based activity in all three regions but greater coupling between regions compared to controls. These collective results may reflect patients’ ability to process basic semantic information but difficulty in selecting correct features, which requires efficient communication between network nodes. This work provides an essential foundation for future investigations aiming to leverage connectomics in improving our understanding of recovery in PWA.

Topic Area: Language Disorders

Back to Poster Schedule