Poster No:
1239
Submission Type:
Abstract Submission
Authors:
Ahmad Samara1, Zaid Zada2, Uri Hasson2, Tamara Vanderwal1, Sam Nastase2
Institutions:
1University of British Columbia, Vancouver, BC, 2Princeton University, Princeton, NJ
First Author:
Co-Author(s):
Introduction:
The brain dynamically reconfigures network connectivity in response to events in the world (1,2). Recent work has used dimensionality reduction to distill this complex, global network structure into principal axes of variance called gradients (3). However, these methods are typically applied to intrinsic connectivity at rest and are entirely data-driven, making it difficult to understand how gradients relate to cortical information processing.
In this study, we develop a modeling framework for interpreting gradients using naturalistic stimuli. First, we use inter-subject functional connectivity (ISFC) to isolate the stimulus-driven network structure that is shared across subjects (4). We then use encoding models to constrain the network structure underlying gradient organization to patterns captured by an explicit model (5,6). Finally, we project these stimulus-driven connectivity patterns back onto the original gradients to quantify which features drive each gradient.
Methods:
We used fMRI data from the Narratives collection (7) in which 46 subjects listened to two different naturalistic narratives. Data were preprocessed using fMRIPrep and vertex-wise time series were averaged within 1000 cortical parcels based on a functional atlas (8). For each story, we extracted three different kinds of language features: acoustic features, syntactic dependencies, and contextual embeddings from the large language model Gemma (as well as confounds capturing word onsets/offsets).
We used banded ridge regression (9) to estimate joint parcel-wise encoding models combining the confound, acoustic, syntactic, and contextual features from a training story, and generated model-based predictions for each feature band for a held-out test story. We evaluated these model-based predictions across subjects and parcels to construct model-based ISFC (mISFC) matrices. We then compared these mISFC matrices to both within-subject function connectivity (WSFC) and inter-subject functional connectivity (ISFC) matrices computed from the same data.
Next, we applied PCA to the group-averaged WSFC matrix to obtain gradients during story-listening. The ISFC and mISFC matrices were projected onto the gradient axes learned from WSFC. We then systematically examined how much variance in connectivity from the ISFC and mISFC matrices aligned with each WSFC gradient. This analysis was replicated with two additional constraints: (1) all input matrices were reduced to include only parcels with strong inter-subject correlation (ISC); (2) we estimated the gradients from ISFC matrices rather than WSFC.
Results:
Story-listening WSFC gradients differed from previously described rest and movie gradients but were highly similar across stories (10). ISFC and mISFC variance was distributed across several gradients with larger proportions of variance aligning with specific gradients. Overall, the amount of variance along each gradient was stratified by model complexity: the contextual embeddings model yielded significantly more variance in connectivity aligned with the top gradients than the lower-order acoustic and syntactic models. This finding was accentuated in the analysis limited to parcels with strong ISC. We also qualitatively examined brain maps for gradients that more closely aligned with ISFC and mISFC. For example, Gradient 2 was anchored by auditory/language areas, suggesting that some gradients are more aligned with stimulus linguistic features. Notably , gradients obtained by applying PCA to ISFC differed from WSFC gradients and were less consistent across stories.
Conclusions:
Our findings demonstrate that connectivity gradients estimated from WSFC are largely driven by intrinsic brain states and are not closely aligned to stimulus-dependent network structure or particular stimulus features. Nonetheless, the large language model recapitulated more of the gradient structure than low-level models and highlighted specific gradients that are largely driven by stimulus-dependent connectivity patterns.
Language:
Language Comprehension and Semantics 2
Speech Perception
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural) 1
fMRI Connectivity and Network Modeling
Methods Development
Keywords:
Cortex
FUNCTIONAL MRI
Language
Modeling
Other - Naturalistic, Gradients
1|2Indicates the priority used for review
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
I accept
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information.
Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
I do not want to participate in the reproducibility challenge.
Please indicate below if your study was a "resting state" or "task-activation” study.
Task-activation
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Healthy subjects
Was this research conducted in the United States?
Yes
Are you Internal Review Board (IRB) certified?
Please note: Failure to have IRB, if applicable will lead to automatic rejection of abstract.
Not applicable
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel?
NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Yes
Were any animal research approved by the relevant IACUC or other animal research panel?
NOTE: Any animal studies without IACUC approval will be automatically rejected.
Not applicable
Please indicate which methods were used in your research:
Functional MRI
Computational modeling
For human MRI, what field strength scanner do you use?
3.0T
Which processing packages did you use for your study?
Other, Please list
-
fMRIPrep
Provide references using APA citation style.
1. Bassett, D. S., & Sporns, O. (2017). Network neuroscience. Nature Neuroscience, 20(3), 353–364.
2. Buzsaki, G. (2019). The Brain from Inside Out. Oxford University Press.
3. Margulies, D. S., Ghosh, S. S., Goulas, A., Falkiewicz, M., Huntenburg, J. M., Langs, G., ... & Smallwood, J. (2016). Situating the default-mode network along a principal gradient of macroscale cortical organization. Proceedings of the National Academy of Sciences, 113(44), 12574–12579.
4. Simony, E., Honey, C. J., Chen, J., Lositsky, O., Yeshurun, Y., Wiesel, A., & Hasson, U. (2016). Dynamic reconfiguration of the default mode network during narrative comprehension. Nature Communications, 7, 12141.
5. Huth, A. G., De Heer, W. A., Griffiths, T. L., Theunissen, F. E., & Gallant, J. L. (2016). Natural speech reveals the semantic maps that tile human cerebral cortex. Nature, 532(7600), 453–458.
6. Zada, Z., Goldstein, A., Michelmann, S., Simony, E., Price, A., Hasenfratz, L., ... & Hasson, U. (2024). A shared model-based linguistic space for transmitting our thoughts from brain to brain in natural conversations. Neuron, 112(18), 3211–3222.
7. Nastase, S. A., Liu, Y. F., Hillman, H., Zadbood, A., Hasenfratz, L., Keshavarzian, N., ... & Hasson, U. (2021). The “Narratives” fMRI dataset for evaluating models of naturalistic language comprehension. Scientific Data, 8(1), 250.
8. Schaefer, A., Kong, R., Gordon, E. M., Laumann, T. O., Zuo, X. N., Holmes, A. J., ... & Yeo, B. T. (2018). Local-global parcellation of the human cerebral cortex from intrinsic functional connectivity MRI. Cerebral Cortex, 28(9), 3095–3114.
9. Dupré la Tour, T., Eickenberg, M., Nunez-Elizalde, A. O., & Gallant, J. L. (2022). Feature-space selection with banded ridge regression. NeuroImage, 264, 119728.
10. Samara, A., Eilbott, J., Margulies, D. S., Xu, T., & Vanderwal, T. (2023). Cortical gradients during naturalistic processing are hierarchical and modality-specific. NeuroImage, 271, 120023.
No