Integrating Subcortical Structures into Whole Brain Models

47 concepts42 papers2 proposals

For decades, cognitive and computational neuroscience has operated under an implicit assumption: to understand human intelligence, perception, and psychiatric disease, one must look to the cerebral cortex. This massive, highly folded sheet of neurons is responsible for the complex hierarchies of prediction, language, and abstract thought. However, this intense focus has led to what Peter Dayan termed "cortical chauvinism" and Frederic Alexandre described as "corticocentric myopia." By treating the cortex as an isolated system, traditional models ignore the evolutionarily older, deep-brain structures—the thalamus, basal ganglia, amygdala, and cerebellum—that act as the brain's vital engine.

A growing movement across the field is working to dismantle this bias. By integrating subcortical structures into whole-brain frameworks, researchers are discovering that the cortex cannot be understood in isolation; its dynamics, stability, and behavioral outputs are fundamentally governed by its deep-brain partners.

Why the Cortex Cannot Act Alone

To understand why a purely cortical model fails, consider how the brain processes a simple sensory-to-motor transformation, such as reacting to an environmental cue. The cortex can build a predictive model of sensory input, but it cannot execute an action or gate information on its own.

As Peter Dayan argues in Palmistry, cortical hierarchies lack the subcortical machinery necessary for value-based decision-making and precise motor control. The translation of high-level sensory predictions into goal-directed actions relies inherently on subcortical systems like the basal ganglia and cerebellum, a mechanism demonstrated computationally by Haroon Anwar et al. in Training a spiking neuronal network model of visual-motor cortex to play a virtual racket-ball game using reinforcement learning.

Furthermore, information does not simply flow freely across distant cortical regions. It is actively modulated, amplified, and gated by subcortical loops:

The Technical Roots of the Exclusion

If subcortical integration is so critical, why have models historically excluded it? The bias is largely driven by two technical constraints: signal-to-noise ratios (SNR) and geometric representation.

First, deep brain structures are notoriously difficult to image using non-invasive methods like fMRI and EEG. As Lucina Q. Uddin et al. point out in Controversies and progress on standardization of large-scale brain network nomenclature, signal-to-noise challenges in functional imaging have led network taxonomies to systematically omit the thalamus and cerebellum. This sentiment is echoed by Huaning Wang et al. in Evidence of a dissociation pattern in default mode subnetwork functional connectivity in schizophrenia, noting that low SNR in deep structures often forces researchers to exclude them from functional connectivity models, blinding us to their role in neurological conditions.

Second, modern neuroimaging has heavily favored surface-based analysis, which inflates and flattens the highly folded cerebral cortex into 2D coordinate systems. While elegant, Gia H. Ngo et al. (From Connectomic to Task-Evoked Fingerprints: Individualized Prediction of Task Contrasts from Resting-State Functional Connectivity) and Simon Dahan et al. (SIM: Surface-based fMRI Analysis for Inter-Subject Multimodal Decoding from Movie-Watching Experiments) emphasize that this surface-centric approach inherently omits the volumetric, nuclear geometry of subcortical anatomy. This prevents models from capturing the full pathway of information flow from initial perception to higher-level cortical processing.

Building a True Whole-Brain Architecture

To overcome these limitations, recent work focuses on developing frameworks that treat the cortex and subcortex as an inseparable, integrated system. This integration is manifesting in three key areas:

1. Developmental Mapping

Rather than focusing solely on cortical expansion, researchers are mapping the co-maturation of deep structures. Liangjun Chen et al. developed a 4D spatiotemporal atlas of the infant subcortex in Four-dimensional mapping of dynamic longitudinal brain subcortical development and early learning functions in infants, demonstrating that the growth of these internal regions directly predicts the emergence of language and motor skills, proving they are essential functional partners during early development.

2. Unified Network Taxonomies

To move past purely cortico-centric parcellations, new computational tools are incorporating subcortical volumes. Ru Kong introduced the Network Correspondence Toolbox (A network correspondence toolbox for quantitative evaluation of novel neuroimaging results) to provide standardized quantitative evaluations for subcortical and cerebellar structures. Similarly, Mo Wang developed the Deep Cluster Atlas framework (DCA: Graph-Guided Deep Embedding Clustering for Brain Atlases) to generate personalized maps that include both cortical and subcortical white-matter structures, improving functional homogeneity.

3. Dynamic Network Reconfiguration

Incorporating subcortical nodes fundamentally changes how we understand cognitive state transitions. Michael W. Cole et al. argue in The Functional Relevance of Task-State Functional Connectivity that network models focusing exclusively on cortical-to-cortical interactions limit our understanding of how the brain reconfigures during tasks. Including the thalamus and basal ganglia is necessary to explain how the brain dynamically shifts its network structure to support cognitive demands.

By shifting from a cortico-centric view to an integrated, loop-based architecture, neuroscience is moving closer to models that don't just simulate isolated cognitive patterns, but capture the full, embodied control loop of the central nervous system.

Key Foundational Work

Palmistry (2004) — contributed 2 concept elements to this direction.
Altered Modular Organization of Structural Cortical Networks in Children with Autism (2013) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_cortical_subcortical_integration_d144bb71,
  title = {cortical-subcortical integration},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/cortical-subcortical-integration/rdr_d144bb71fbe04985aaba2e50224fc38c},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_cortical_subcortical_integration_d144bb71_v1,
  title = {cortical-subcortical integration},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/cortical-subcortical-integration/rdr_d144bb71fbe04985aaba2e50224fc38c/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake