The Intrinsic Manifold of Spontaneous Activity Constrains Cortical Responses to Naturalistic Stimuli
Preprint
Abstract
The cerebral cortex constrains its spontaneous activity to a low-dimensional manifold learnable from resting state functional magnetic resonance imaging (fMRI) data. However, it remains unclear whether this intrinsic manifold also captures cortical responses to complex, naturalistic stimuli. To test this, we pretrained a deep variational autoencoder model on 3-Tesla resting-state fMRI data to learn the latent structure of spontaneous activity, and then applied this model without finetuning to 7-Tesla fMRI data acquired during movie-watching. Despite the different field strengths, the model generalized robustly from resting to movie-watching states. The latent representation of stimulus-evoked responses was confined to a subspace that occupied about 13% of the latent space spanned by spontaneous activity, demonstrating that task-related neural responses do not require a distinct representational space. By representing cortical dynamics as an evolving latent trajectory, we found striking differences across individuals or between brain states. During movie watching, the velocity of the latent trajectory provided a reliable marker of cortical engagement, and its temporal structure was highly reliable and sensitive to naturalistic events. These findings suggest that the intrinsic manifold of spontaneous activity forms a full reservoir of cortical states that the brain can differentially engage when interacting with the external environment.Keywords
Funding Information
- National Institute on Aging (AG082204)
This publication has 36 references indexed in Scilit:
- Naturalistic stimuli: A paradigm for multiscale functional characterization of the human brainCurrent Opinion in Biomedical Engineering, 2021
- Understanding Heterogeneity in Clinical Cohorts Using Normative Models: Beyond Case-Control StudiesBiological Psychiatry, 2016
- Automatic denoising of functional MRI data: Combining independent component analysis and hierarchical fusion of classifiersNeuroImage, 2014
- The WU-Minn Human Connectome Project: An overviewNeuroImage, 2013
- Pushing spatial and temporal resolution for functional and diffusion MRI in the Human Connectome ProjectNeuroImage, 2013
- The minimal preprocessing pipelines for the Human Connectome ProjectNeuroImage, 2013
- Modeling the hemodynamic response function in fMRI: Efficiency, bias and mis-modelingNeuroImage, 2009
- Advances in functional and structural MR image analysis and implementation as FSLNeuroImage, 2004
- Functional connectivity in the motor cortex of resting human brain using echo‐planar mriMagnetic Resonance in Medicine, 1995
- Statistical parametric maps in functional imaging: A general linear approachHuman Brain Mapping, 1994