Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Poster Slams
Univariate and multivariate mapping of the brain regions involved in word retrieval across stroke and neurodegeneration
Poster C39 in Poster Session C, Friday, October 7, 10:15 am - 12:00 pm EDT, Millennium Hall
Diego Lorca-Puls1,2, Thomas Hope3, Andrea Gajardo-Vidal1,4, Maria Luisa Mandelli1, Rian Bogley1, Abigail Licata1, PLORAS Team3, Alexander Leff3, David Green3, Nina Dronkers5,6, Cathy Price3, Maria Luisa Gorno-Tempini1; 1University of California San Francisco, 2Universidad de Concepción, 3University College London, 4Universidad del Desarrollo, 5University of California Berkeley, 6University of California Davis
The ability to retrieve a word one wants to say is a fundamental human skill that frequently breaks down after brain injury. Behaviorally, it is typically measured by having patients name the objects depicted in a sequence of pictures. Prior lesion studies of word retrieval have implicated a wide variety of discrete brain regions, often in an around the left temporal lobe. One important factor that arguably contributed to this variability in the mapping between lesion and deficit is the spatial bias, and ensuing localization errors, that follow from the non-random distribution of brain damage: e.g., stroke-induced brain lesions are constrained by the anatomy of the vascular tree, while neurodegenerative diseases affect (preferentially) vulnerable large-scale neural networks. Critically, lesion-deficit relationships observed across different etiologies with largely uncorrelated spatial biases would be much less likely to be contaminated by localization errors. Here, we embrace this logic to localize brain regions that are persistently necessary for successful word retrieval in two large samples of patients suffering from brain damage caused by stroke (n = 587) or neurodegeneration (n = 205). To further enhance the robustness of our results, we combined both mass-univariate and multivariate voxel-based lesion-deficit mapping approaches, as implemented by the SVR-LSM toolbox (https://github.com/atdemarco/svrlsmgui) using default settings, including cluster-level family-wise error correction via permutation testing. Moreover, to ensure the identification of brain regions involved in word retrieval specifically, we considered a set of nuisance covariates to control for components of object naming related to, for example, visual perception, semantic processing, phonological encoding and articulation; in addition to accounting for the influence of time since onset and lesion volume, as is routinely done in lesion-deficit analysis. We identified a region in the left mid-to-anterior superior temporal sulcus/middle temporal gyrus, extending into neighboring white matter, which was consistently associated with word retrieval ability across both analysis types (univariate and multivariate) and neurological conditions (stroke and neurodegeneration). We also show, in an entirely independent sample of stroke patients (n = 236), that damage in this region predicts (i.e., out-of-sample) around a third of the variance in object naming scores, underscoring its scientific and clinical importance. Taken together, our findings strongly imply that we have accurately localized a left temporal region that is a critical node in the object naming network, and where damage is likely to impair word retrieval ability.
Topic Areas: Disorders: Acquired, Meaning: Lexical Semantics