Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Lightning Talks
cneuromod-triplets: enabling deep lexico-semantic phenotyping through a large and freely available dataset
There is a Poster PDF for this presentation, but you must be a current member or registered to attend SNL 2023 to view it. Please go to your Account Home page to register.
Poster E113 in Poster Session E, Thursday, October 26, 10:15 am - 12:00 pm CEST, Espace Vieux-Port
Abigail E. Licata1, Pierre Bellec2,3, Simona Brambati2,3, Valentina Borghesani1; 1Université de Genève, 2Centre de recherche de l'Institut universitaire de gériatrie de Montréal, 3Université de Montréal
Existing neurocognitive models of lexico-semantic processing describe a broad, bilateral fronto-temporal network. However, the role of individual variability has often been overlooked and can be best examined through deep phenotyping (Gratton & Braga, 2021): the collection, within subject, of behavioral and neuroimaging responses to stimuli spanning a wide range of variables during tasks that require different levels of semantic processing. We present the cneuromod-triplets dataset from the Courtois Project on Neural Modelling (cneuromod, https://www.cneuromod.ca/): a large-scale, freely available and multimodal dataset that covers a wide range of neuroimaging and behavioral tasks in a small sample of subjects. The dataset includes two tasks that enable 1) the isolation of neural activity associated to single word processing and its modulation by perceptual and conceptual features of words and 2) the investigation of semantic processing of those same words within a minimal context. The first task was a familiarity judgment task in which subjects were visually presented with a single word and were asked to rate their subjective familiarity (Fernandino et al., 2022). The second task was a three-term task (3TT) in which participants were presented with three words (i.e., a "triplet") and asked to select the unrelated word from the set. The stimuli, drawn from a large, publicly available and human-annotated dataset (Borghesani et al., 2023), comprised 1,588 unique words spanning a large range of lexico-semantic and psycholinguistic features, arranged into 709 triplets. Four right-handed, English-French bilingual participants (2 female, age: 44±4.3 years; education: 20±3.6 years) performed these two tasks in a 3T fMRI scanner over multiple sessions, for a total of approximately seven hours. To highlight the quality of this dataset, we ran parametric modulation analyses for each subject to examine fMRI-BOLD response covariation with low-level visual (i.e., word length), lexical (i.e., frequency, orthographic complexity) and semantic (i.e., concreteness, sensorimotor features) aspects of the presented single words, as well as the effect of response time (a proxy for task difficulty) during the 3TT. Lexico-semantic features of single words differentially modulated neural activity across the cortex, with frequency and sensorimotor features of the words showing the greatest and most heterogeneous modulatory pattern. Our results highlight the quality of this freely available dataset, which enables further investigation of the organization of lexico-semantic processing at the individual level. For instance, we will leverage multivariate methods to assess whether the subject-specific trial-by-trial decision on the 3TT can be predicted by the pattern of activation of the three words when presented individually during the familiarity task (Wu et al., 2022).
Topic Areas: Development of Resources, Software, Educational Materials, etc., Meaning: Lexical Semantics