Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Cortical sites critical to speech and language function have distinct static and dynamic network profiles
Poster Session A, Friday, September 12, 11:00 am - 12:30 pm, Field House
Marc Slutzky1, Robert Flint1, Jason Hsieh1, Prashanth Prakash1, Nathan Crone2, Joshua Rosenow1, Todd Parrish1, Matthew Tate1, Richard Betzel3, Jeremy Greenlee4; 1Northwestern University, 2Johns Hopkins University, 3University of Minnesota, 4University of Iowa
Prior to surgery for epilepsy or brain tumors, focal brain areas that are “critical” for speech and language have long been identified using electrocortical stimulation (ECS), which perturbs function. However, the precise mechanisms of ECS are poorly understood, despite its established utility in the clinic. For example: speech/language circuits in the brain are recognized to span multiple cortical areas, but it remains unclear whether ECS affects only the local cortex, or if network-level processing is disrupted. To investigate this question, we recorded electrocorticography (ECoG) from sixteen participants while they performed a single-word reading aloud task. We extracted high-gamma (70-200 Hz) activity from ECoG recordings from frontal, temporal, and parietal electrodes, and quantified the properties of cortical networks as they relate to both criticality and behavior. To do this, we first computed pairwise correlations between electrodes over different timescales—static (several minutes) and dynamic (750-ms windows sliding by 100-ms increments) to examine changes in network properties during behavior. We quantified network properties using graph theory metrics computed from the correlations. As noted, we were particularly interested in the static and dynamic properties of critical sites for speech and language (those producing speech arrests and language errors such as anomias and paraphasias, respectively) compared to other (non-critical) cortical sites. Static analysis showed that critical sites could be distinguished by their profile of network properties. This profile included lower global and local connectivity (clustering coefficient, eigenvector centrality) for both speech arrest and language error sites than for non-critical sites. Notably, language error sites served as connectors between modules (subnetworks) in the language network (i.e., had high participation coefficients). Additionally, cortical networks reconfigured dynamically during oral word-reading. Transitions between behavioral states—for example, among rest, task engagement, and speech onset—changed the relationships between measures of local connectivity and connectivity across subnetworks. In particular, there was a change in the correlation between participation coefficient (“connectorness”) and within-module degree (local connectivity) for critical sites, but not for non-critical sites. These patterns of change meant that, broadly, cortical network connectivity during the behavior could deviate substantially from its profile as calculated with a static analysis. These results suggest that a site’s pattern of connections within the language network helps determine its importance to language function. Further, language networks’ properties change dynamically and heterogeneously with behavior. This highlights the importance of examining brain activity during active behavior, at the resolution of the behavior itself, in addition to static resting state analysis. It also suggests that sites critical for language and speech function can be defined by dynamic, quickly changing connectivity profiles.
Topic Areas: Language Production, Computational Approaches