Abstract
Encoding models for mapping voxelwise semantic tuning are typically estimated separately for each individual, limiting their generalizability. In the current report, we develop a method for estimating semantic encoding models that generalize across individuals. Functional MRI was used to measure brain responses while participants freely viewed a naturalistic audiovisual movie. Word embeddings capturing agent-, action-, object-, and scene-related semantic content were assigned to each imaging volume based on an annotation of the film. We constructed both conventional within-subject semantic encoding models and between-subject models where the model was trained on a subset of participants and validated on a left-out participant. Between-subject models were trained using cortical surface-based anatomical normalization or surface-based whole-cortex hyperalignment. We used hyperalignment to project group data into an individual’s unique anatomical space via a common representational space, thus leveraging a larger volume of data for out-of-sample prediction while preserving the individual’s fine-grained functional–anatomical idiosyncrasies. Our findings demonstrate that anatomical normalization degrades the spatial specificity of between-subject encoding models relative to within-subject models. Hyperalignment, on the other hand, recovers the spatial specificity of semantic tuning lost during anatomical normalization, and yields model performance exceeding that of within-subject models.
Highlights
Recent neuroimaging work has revealed widespread cortical representation of semantic content conveyed by visual and linguistic stimuli (Huth et al, 2012, 2016; Wehbe et al, 2014; Pereira et al, 2018)
To ensure that the common space learned by hyperalignment finds common bases for fine-grained functional topographies across subjects, we computed ISCs for both vertex-wise response time series and searchlight representational geometries using anatomical normalization and hyperalignment
We developed a framework for constructing between-subject semantic encoding models that generalize to both novel stimuli and novel subjects
Summary
Recent neuroimaging work has revealed widespread cortical representation of semantic content conveyed by visual and linguistic stimuli (Huth et al, 2012, 2016; Wehbe et al, 2014; Pereira et al, 2018) These findings hinge on the development of forward encoding models, which find a mapping from stimuli to voxelwise responses via a complex intermediate feature space (Naselaris et al, 2011). Güçlü and van Gerven (2017) and Wen et al (2018) have employed hyperalignment in conjunction with a deep convolutional neural network (e.g., Tran et al, 2015) to predict responses to video clips visual areas They demonstrated that estimating an encoding model in a common representational space does not diminish model performance, and that aggregating additional subjects in the common spaces can improve performance. Hyperalignment effectively recovers the specificity of within-subject models, allowing us to leverage a large volume of group data for individualized prediction at the specificity of individual voxels or cortical vertices
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.