Whole-brain fMRI reveals the notes, chords, and conductors of the cortical orchestra
Poster No:
2100
Submission Type:
Abstract Submission
Authors:
Brandon Munn1, James Shine2
Institutions:
1University of Sydney, Sydney, NSW, 2University of Sydney, Sydney, NA
First Author:
Co-Author:
Introduction:
The brain produces rich and dynamic patterns of neural activity spanning multiple spatial and temporal scales over seconds to minutes and in circuits and systems to enable a broad spectrum of complex behaviours. How does the brain instantiate this wide variety of spatiotemporal dynamics? A prevailing assumption in the field, which is currently undergoing revision, is that these features arise due primarily to the organisation of and the interactions within the cerebral cortex1. In this study, we demonstrate that many features of coordinated neural activity measured by standard whole-brain functional magnetic resonance imaging (fMRI) approaches are inherently linked to the organisation of the subcortex (e.g., thalamus, neuromodulatory brainstem structures, cerebellum – Fig. 1a). Leveraging a geometrically-based eigenmode decomposition of (i.e., the "notes" of the brain2; Fig. 1b) and a temporal dimensionality reduction technique, identifying time-lagged independent components (i.e., the "chords", which are different combinations of notes – Fig. 1c), we establish a connection between slow subcortical fMRI temporal dynamics with spatially coarse cortical patterns and vice versa for fine spatial patterns.
Methods:
7T fMRI dataset: We analysed a dataset of 8 participants who undertook multiple fMRI scans9 consisting of alternating resting-state with visually evoked scans viewing an extensive set (~10k) of naturalistic images and responding to a behavioural recognition task. Importantly, this dataset was obtained using a 7T scanner at 1.8 mm isotropic resolution (TR = 1.6s)6.
Low-dimensional mapping between subcortical activity and cortical modes: We extracted cortical activity into natural oscillatory modes of the physical cortex based (i.e., its spatial eigenmodes), analogous to the notes of a stringed instrument, represent the optimal basis set for the systematic decomposition of cortical neural activity2,10. We then applied time-lagged independent component analysis (tICA) to the subcortical ROIs and cortical modes. tICA combines information from a time-lagged covariance matrix of the data to separate blood oxygen level-dependent activity into a nonlinear mixture of dynamic whole-brain modes.
Low-dimensional mapping between subcortical activity and cortical modes: We extracted cortical activity into natural oscillatory modes of the physical cortex based (i.e., its spatial eigenmodes), analogous to the notes of a stringed instrument, represent the optimal basis set for the systematic decomposition of cortical neural activity2,10. We then applied time-lagged independent component analysis (tICA) to the subcortical ROIs and cortical modes. tICA combines information from a time-lagged covariance matrix of the data to separate blood oxygen level-dependent activity into a nonlinear mixture of dynamic whole-brain modes.
Results:
Our methodology reveals that activation in subcortical structures significantly precede resting-state component time series in the cerebral cortex (explained variance > 0.6). These patterns further explain the well-known alternating oscillation in dorsal attention and default mode networks (Fig. 1d). In particular, we focus on the anatomical role of these systems, as we find that diffuse projecting thalamic and neuromodulatory systems preferentially explain slow and coarse spatiotemporal patterns. This extension can be seen as a whole-brain extension to previous theoretical3–5 and empirical6–8 analyses focusing on one subcortical structure at a time. In particular, we confirm that the independent spatiotemporal whole- brain modes typically involve a coordinated orchestra of subcortical structures.
Conclusions:
Our study relates how multiscale geometric cortical eigenmodes (i.e., notes of the brain) are coordinated into dynamic motifs (i.e., Chords), that are explained by subcortical activity (the conductors). This approach directly contributes to systems neuroscience by functionally linking whole-brain systems and delineating how these patterns align with subcortical anatomical connections. Nevertheless, we believe that incorporating coarse-grained subcortical systems dynamics will significantly enhance the accuracy of whole-brain computational models, and modelling will be required to understand the complex dynamical interplay of these various structures.
Modeling and Analysis Methods:
Multivariate Approaches 2
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Anatomy and Functional Systems 1
Novel Imaging Acquisition Methods:
BOLD fMRI
Keywords:
Acetylcholine
Adrenaline
Brainstem
HIGH FIELD MR
Multivariate
Open Data
Sub-Cortical
Thalamus
1|2Indicates the priority used for review