Optimal signal transmission and timescale diversity in a model of human brain operating near criticality
Yang Qi · 2024
Read Optimal signal transmission and timescale diversity in a model of human brain operating near criticality by Yang Qi (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-17043.txt.
Introduction
Yang Qi and co-authors model how the brain passes signals and varies in speed in 2024. Brain areas answer inputs at many speeds, from fast to slow. That spread is thought to help the brain handle events on many time scales. A working brain must keep that spread while still passing signals without heavy loss.
The team built a simplified model from a digital twin of the human brain. The twin carries the full map of links between brain regions. Tests show link strength across regions shapes both speed spread and signal flow. Running near a critical point may let the brain gain both traits at once.
Catalog record
| CREATOR | Yang Qi |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 79 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-17043
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Optimal signal transmission and timescale diversity in a model of human brain operating near criticality Yang Qi Abstract Abstract Cortical neurons exhibit a hierarchy of timescales across brain regions in response to input stimuli, which is thought to be crucial for information processing of different temporal scales. Modeling studies suggest that both intra-regional circuit dynamics as well as cross-regional connectome may contribute to this timescale diversity. Equally important to diverse timescales is the ability to transmit sensory signals reliably across the whole brain. Therefore, the brain must be able to generate diverse timescales while simultaneously minimizing signal attenuation. To understand the dynamical mechanism behind these phenomena, we develop a second-order mean field model of the human brain by applying moment closure and coarse-graining to a digital twin brain model endowed with whole brain structural connectome. Cross-regional coupling strength is found to induced a phase transition from asynchronous activity to synchronous oscillation. By analyzing the input-response properties of the model, we reveal criticality as a unifying mechanism for enabling simultaneously optimal signal transmission and timescales diversity. We show how structural connectome and criticality jointly shape intrinsic timescale hierarchy across the brain. 1 Introduction The human brain is a highly heterogeneous and complex system, with distinct regions serving specialized functions. The coordinated interactions among these regions collectively shape a hierarchical mode of information processing. Recent experimental evidence suggests that neural response to input stimuli exhibits a timescale gradient across the whole brain in both human and non-human primates [ 1 , 2 , 3 , 4 , 5 ] . In particular, sensory areas respond with brief, transient reactions to input stimuli whereas higher-order cortical areas produce responses that remain persistent over longer durations [ 6 , 5 , 7 ] . This hierarchy of intrinsic timescales are thought to play key roles in processing sensory inputs and generating behavior at different temporal scales [ 2 , 6 , 8 , 9 ] . Modeling studies have attributed the emergence of timescale hierarchy to heterogeneity in self-coupling [ 10 ] , local recurrent connectivity [ 11 , 12 ] , and whole brain anatomical connectome [ 13 , 14 ] . Notably, computational modeling of the monkey brain has revealed that the intrinsic timescales gradually increase from area to area along the cortical hierarchy [ 13 , 14 ] . Using linear network models, it has been shown that intrinsic timescales become segregated if the eigenmodes of the connectivity matrix are localized to different parts of the network [ 11 , 13 ] . Such localization of eigenmodes may arise from heterogeneity in spatially local connectivity [ 11 ] as well as connectivity endowed with anatomical connectome [ 13 ] . Despite these theoretical insights, how intrinsic neural timescales may arise in the human brain has yet been explored. Although previous modeling studies have focused on explaining the origin of timescale hierarchies in the brain, the signal transmission aspect of input processing has not been fully addressed. If input signals attenuate too quickly as it propagates across the cortex, then responses in higher order cortical areas with long timescales may become negligible. This raises the question how the brain might attain diverse timescales while simultaneously ensuring reliable signal transmission. In nature, many physical systems operate in a regime that exhibits signatures of criticality, which is thought to be optimal for signal transmission [ 15 , 16 ] . This has inspired the critical brain hypothesis which suggests that the brain may gain significant information-processing advantages by operating near the critical point of a phase transition, including heightened sensitivity to perturbations, a diverse range of system states, maximal brain fluidity, and increased capacity for information storage and transmission [ 17 , 18 , 19 , 20 , 21 , 22 , 12 ] . In this study, we develop a second-order dynamic mean field model of the human brain by applying moment closure and coarse-graining to a recently proposed Digital Twin Brain (DTB) model for simulating the whole brain at single neuron resolution [ 23 , 24 ] . By globally tuning the strength of synaptic connections between brain regions, we show that the mean field model exhibits a phase transition from asynchronous irregular activity to brain-wide synchronous oscillations, as consistent with the DTB model. Using the mean field model, we investigate the processing of visual sensory signals across the whole brain, with special focuses on the strength of signal transmitted to each brain region and the hierarchy of intrinsic timescales as the system state approaches criticality. 2 Results 2.1 Moment closure, coarse-graining, and a mean field model for the human brain To link whole brain dynamics at the macroscale to biophysical nonlinearity at the microscale, we develop a multiscale approach to whole brain modeling, as illustrated schematically in Fig. 1 (A). We begin with a region-wise computational cortico-subcortical model of whole human brain at single neuron resolution [ 23 , 24 ] . The model comprises n=378 cortical and subcortical regions according to the brain parcellation HCPex [ 25 ] . Each area is represented by a conductance-based spiking neural circuit containing one excitatory (AMPA-ergic) population and one inhibitory (GABA-ergic) population. Different brain regions are connected only by synaptic projections between excitatory neurons. Each neuron is modeled using a conductance-based leaky integrate-and-fire neuron with finite synaptic time constants. The synaptic connectivity within and between regions are drawn randomly according to certain distributions with given average in-degrees. The intra-regional and the cross-regional synaptic in-degrees, the conductance of each region, and the number of neurons within each brain region are estimated based on biological constraints (see Methods). External inputs to different brain regions are determined via a data assimilation approach [ 23 , 24 ] . A global scaling factor \gamma controls the cross-regional synaptic coupling strength. We apply moment closure and coarse-graining to arrive at a second-order mean field model for the human brain. The moment closure is performed on the spiking neural circuit model through a Fokker-Planck formalism. The coarse-graining is performed by treating each brain region as a homogeneous population of neurons with identical firing statistics. Under these treatments, the state vector \mathbf{m} of the mean field model is \mathbf{m}=(\boldsymbol{\mu}_{E},\boldsymbol{\sigma}^{2}_{E},\boldsymbol{\mu}_{I},\boldsymbol{\sigma}^{2}_{I})\in\mathbb{R}^{4n} , where n=378 is the number of brain regions. Here, \boldsymbol{\mu}_{\alpha}=(\mu^{\alpha}_{u}) and \boldsymbol{\sigma}^{2}_{\alpha}=(\sigma^{\alpha,2}_{u}) represent the mean firing rate and firing variability of neural population \alpha\in\{E,I\} in region u . We can then write down the mean field model as \displaystyle\tau\dfrac{d}{dt}\left(\begin{array}[]{c}\boldsymbol{\mu}_{E}\\ \boldsymbol{\sigma}^{2}_{E}\\ \boldsymbol{\mu}_{I}\\ \boldsymbol{\sigma}^{2}_{I}\\ \end{array}\right)=-\left(\begin{array}[]{c}\boldsymbol{\mu}_{E}\\ \boldsymbol{\sigma}^{2}_{E}\\ \boldsymbol{\mu}_{I}\\ \boldsymbol{\sigma}^{2}_{I}\\ \end{array}\right)+\Phi\left(\begin{array}[]{c}\boldsymbol{\mu}_{E}\\ \boldsymbol{\sigma}^{2}_{E}\\ \boldsymbol{\mu}_{I}\\ \boldsymbol{\sigma}^{2}_{I}\\ \end{array}\right). (1) The time constant \tau is determined to be approximately 11 ms by fitting the oscillation frequency of the mean field model to that of the spiking model. The coarse-grained moment mapping \Phi can be decomposed into \Phi=\phi_{\rm{MA}}\circ\phi_{\rm{eff}}\circ\phi_{\rm{sum}}. (2) The first component is the coarse-grained synaptic summation \phi_{\rm sum}:\mathbb{R}^{4n}\to\mathbb{R}^{8n} which maps the moments of pre-synaptic spike count to that of synaptic activations at post-synaptic neurons. The second component is the effective current \phi_{\rm eff}:\mathbb{R}^{8n}\to\mathbb{R}^{6n} , whose output includes an effective time constant and the moments of the effective current. The third component \phi_{\rm MA}:\mathbb{R}^{6n}\to\mathbb{R}^{4n} is the moment activation providing the moments of the output spike count. See Methods for details of these moment mappings. 2.2 Phase transition induced by