Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Effective localization of high-level language with language-responsive fROIs
Poster Session C, Thursday, October 1, 10:45 am - 12:45 pm, Wangari Maathai
M. Ryan Henderson1, Rebecca Belisle1, Yinuo Liu1, Terri Scott2, Tyler Perrachione1; 1Boston University, 2Northeastern University
Recent advances in functional neuroimaging have enabled the identification of areas in individual brains that are maximally selective for high-level language processing using standalone language localizer tasks (Fedorenko et al., 2024). This subject-specific approach has enabled the field to substantively improve the theoretical precision and statistical power of neuroimaging experiments by controlling for variability in the spatial realization of language areas across individuals. However, it is unclear whether localizing language areas always requires a “cognitive subtraction” between a language task and a control task, or whether language-selective regions can be localized using a simpler metric of neural responsiveness to language over resting baseline. Using the resting baseline in place of a control condition contrast could benefit fMRI language localization in many ways, including reducing localizer scanner time, being more appropriate for special populations who may show different, even preferential, responses to the control condition (Alho et al., 2021), and allowing retrospective subject-specific analyses of datasets collected without a separate localizer. Here, we measured neural activation to a receptive language localizer task that includes blocks of listening to intact, meaningful speech and degraded, unintelligible speech (Scott et al., 2017). We compared the functional properties of subject-specific functional regions of interest (fROIs) defined based on “language responsiveness” (intact speech > rest) versus “language selectivity” (intact speech > degraded speech). We defined subject-specific fROIs as the top 10% of either language-responsive or language-selective voxels within each of 13 probabilistic language parcels obtained from a population-level atlas: bilateral aSTG, pSTG, IFG pars opercularis, IFG pars triangularis, precentral gyrus; left superior frontal gyrus; and right cerebellum (Lipkin et al., 2022). Data independence was preserved by defining and testing fROIs in separate runs. Using linear mixed-effects models, we analyzed whether there were any differences, based on fROI defining contrast, in either language selectivity (intact > degraded speech), task specificity (selectivity for language vs. the multiple-demand task contrasts), and spatial consistency between runs. We found that the contrast used to define the fROIs had no effect on their degree of language selectivity. Bilateral temporal fROIs defined by language responsiveness had overall greater response magnitudes than those defined by selectivity, an effect not found in frontal fROIs. The between-run test-retest reliability of fROI location did not differ by defining contrast. Moreover, the within-run spatial consistency of fROIs defined by each contrast was as good as or better than the fROI consistency across runs. Qualitatively, language-responsive fROIs extended more dorsally on STG, while language-selective fROIs extended more ventrally into STS, but fROI locations did not differ systematically in IFG. Overall, this work demonstrates that voxelwise responsiveness of language vs. resting baseline may be a viable means of determining subject-specific language-selective fROIs in language neuroimaging research.
Topic Areas: Methods,