Related Experiment Video
Updated: Aug 5, 2026
09:44
Recording and Analyzing Multimodal Large-Scale Neuronal Ensemble Dynamics on CMOS-Integrated High-Density Microelectrode Array
Published on: March 8, 2024
Neural manifold connectomics reveals multiregime functional connectivity
Puneet Velidi1, Enrico Amico2,3, Farouk Nathoo1
1Department of Mathematics and Statistics, University of Victoria, 3800 Finnerty Rd., Victoria, V8P 5C2, British Columbia, Canada.
Biorxiv : the Preprint Server for Biology
|August 1, 2026
Summary
This study reveals that the structure of brain activity within regions, not just average signals, contains unique individual information. This richer neural representation improves brain connectivity analysis, especially during cognitive tasks.
Area of Science:
- Neuroscience
- Brain Imaging
- Network Science
Background:
- Functional connectivity in fMRI typically uses averaged time series per brain region, potentially losing detailed neural information.
- Neural activity exhibits low-dimensional structures that are not fully captured by current scalar representations.
Purpose of the Study:
- To explore the information contained within the low-dimensional structure of within-parcel BOLD activity.
- To develop a novel method for quantifying functional connectivity using regional temporal subspaces.
- To investigate how this subspace-based connectivity relates to individual identity and cognitive states.
Main Methods:
- Representing each brain parcel by a low-dimensional temporal subspace derived from principal component time series.
- Quantifying connectivity between regional subspaces using the RV coefficient.
- Analyzing resting-state and working-memory fMRI data from the Human Connectome Project.
Main Results:
- Connectivity derived from the first principal component best identifies individuals at rest.
- Individual identifiability peaks with secondary components included during working memory tasks.
- These findings demonstrate state-dependent information distribution across principal components.
Conclusions:
- Within-parcel BOLD signal structure contains identity- and state-dependent information missed by scalar summaries.
- Functional connectivity can be viewed as a family of connectomes indexed by retained regional subspaces.
- This provides a new framework for analyzing interactions between low-dimensional neural representations.
