Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Language-critical areas serve as connectors across language subnetworks
Poster Session D, Saturday, October 26, 10:30 am - 12:00 pm, Great Hall 3 and 4
Marc Slutzky1, Jason Hsieh1,2, Prashanth Prakash1, Robert Flint1, Emily Mugler1, Yujing Wang3, Nathan Crone3, Jessica Templer1, Joshua Rosenow1, Todd Parrish1, Matthew Tate1, Richard Betzel4, Jeremy Greenlee5; 1Northwestern University, 2Cleveland Clinic Foundation, 3Johns Hopkins University, 4Indiana University, 5University of Iowa
Historically, important abilities such as speech and language were viewed as localized to focal areas of human cerebral cortex. Moreover, direct electrocortical stimulation (ECS) has long been used clinically to identify focal sites thought to be “critical” to speech and language function. This procedure is often performed prior to epilepsy or tumor surgery to avoid damaging these functions from the resection. Yet, more recent studies have shown that large cortical networks are activated during language and speech tasks, leading many to hypothesize that eloquent functions may instead be emergent properties of distributed brain networks. Here, we sought to reconcile these different viewpoints and elucidate the network properties that predict whether cortical sites are labeled by ECS as critical for speech and language. We recorded electrocorticography (ECoG) from sixteen participants who performed a word-reading task while they underwent either awake craniotomy for brain tumor resection, or extraoperative monitoring for epileptogenic focus localization. We extracted high-gamma activity from ECoG recordings and computed the pairwise connectivity. We used modularity maximization to find the community membership of nodes in the task-activated network and calculated several well-described network connectivity metrics. We then used these metrics to examine the network properties of cortical sites defined by ECS to be critical for speech and language (those producing speech arrests and language errors, respectively) and compared to other (non-critical) cortical sites. We discovered different network signatures for cortical sites where ECS caused speech arrest and language errors. Both types of ECS-critical sites were characterized by a lower amount of local (clustering coefficient, local efficiency) and global (eigenvector centrality) connectivity than non-critical network sites. Cortical sites where ECS caused language errors, in particular, exhibited higher participation coefficients — that is, higher inter-community connectivity. Taken together, these metrics constitute a network signature that indicates that language-critical nodes serve as connectors between communities in the language network. Further, we used this set of network features alone to train support vector machine and k-nearest neighbor classifiers to predict which nodes would be critical to speech and language with ECS. These classifiers predicted critical nodes with relatively high accuracy, including across participants. These findings suggest that a site’s pattern of connections within the language network helps determine its importance to language function. For higher-order cognitive functions such as language, which depend on coordinated actions of multiple subnetworks (communities), nodes that connect these subnetworks appear to be critical to function. In contrast, for lower-order functions, such as speech articulation, connectors between communities may be less critical.
Topic Areas: Language Production, Computational Approaches