Source-linked AI summary
Functional Alignment with Anatomical Networks is Associated with Cognitive Flexibility
John D. Medaglia, Weiyu Huang, Elisabeth A. Karuza, Sharon L. Thompson-Schill, Alejandro Ribeiro, Danielle S. Bassett
TL;DR
The paper addresses the lack of a concise measure integrating anatomical and functional organization during cognitive flexibility. Using graph signal processing on diffusion-derived networks and BOLD data from 28 individuals, it finds that stronger functional alignment with anatomical organization is associated with faster cognitive switching, while noting that the cognitive role of highly aligned signals remains unexplained.
Problem
No concise measure integrating anatomical and functional organization in support of cognitive flexibility had been identified.
Method
The study reconstructed individualized anatomical networks from diffusion imaging and decomposed framewise regional BOLD signals by their alignment with white matter network organization.
Results
Greater alignment between functional signals and underlying anatomical network organization was associated with faster cognitive switching across subjects.
Takeaways & Limitations
The approach provides an integrated structure-function perspective on cognitive flexibility and a potential neural biomarker.
Takeaways & Limitations
The results do not explain the potential cognitive role of highly aligned signals.
Abstract
from arXiv · showhide
Cognitive flexibility describes the human ability to switch between modes of mental function to achieve goals. Mental switching is accompanied by transient changes in brain activity, which must occur atop an anatomical architecture that bridges disparate cortical and subcortical regions by underlying white matter tracts. However, an integrated perspective regarding how white matter networks might constrain brain dynamics during cognitive processes requiring flexibility has remained elusive. To address this challenge, we applied emerging tools from graph signal processing to decompose BOLD signals based on diffusion imaging tractography in 28 individuals performing a perceptual task that probed cognitive flexibility. We found that the alignment between functional signals and the architecture of the underlying white matter network was associated with greater cognitive flexibility across subjects. Signals with behaviorally-relevant alignment were concentrated in the basal ganglia and anterior cingulate cortex, consistent with cortico-striatal mechanisms of cognitive flexibility. Importantly, these findings are not accessible to unimodal analyses of functional or anatomical neuroimaging alone. Instead, by taking a generalizable and concise reduction of multimodal neuroimaging data, we uncover an integrated structure-function driver of human behavior.
Article summary
The study integrates anatomical networks and functional signals to examine cognitive flexibility in 28 individuals. Greater alignment between brain signals and white matter network organization was associated with faster, more flexible switching.
- 28 healthy individuals were assessed using cognitive switch costs during functional neuroimaging.
- Individualized white matter networks were reconstructed from diffusion spectrum imaging data.
- Graph signal processing decomposed functional brain signals into components aligned or liberal relative to underlying anatomy.
- Greater alignment between functional signals and anatomical network organization was associated with faster cognitive switching.
- Signals most aligned with the anatomical network were also associated with greater cognitive flexibility.
- The approach provides an anatomically grounded integration of brain structure and function and identifies a potential neural biomarker.
Introduction
The introduction frames cognitive switching as a distributed process requiring integration of regional activity, anatomical connectivity, and behavior. The study addresses this gap by combining diffusion-derived anatomical networks with functional signals during perceptual switching.
- Cognitive flexibility supports switching between attentional foci or processing modalities, but switching incurs measurable response-time costs.
- Cognitive switching engages corticobasal ganglia-thalamo-cortical loops, including anterior cingulate feedback detection and lateral prefrontal rule maintenance.
- Understanding task-switching circuitry remains challenging because regional activity, anatomical connectivity, and behavior must be integrated.
- Existing cognitive-neuroscience and connectomics approaches have largely developed in parallel, while frameworks using white matter structure to constrain functional signals are lacking.
- The study analyzes 28 adults using diffusion spectrum imaging, BOLD fMRI, and a perceptual task switching between local and global shape features.
- Anatomical networks of 111 brain regions were constructed from tractography, and graph eigenspectra decomposed regional BOLD signals by alignment with underlying white matter.
- The approach tests whether moment-to-moment functional alignment supports switching and whether anatomically central subcortical regions have greater functional centrality.
- Unlike region-specific approaches, the method examines local neural processes across the brain’s distributed anatomical network.
associated with faster cognitive switching
Lower liberality, reflecting greater functional alignment with white matter networks, was associated with lower cognitive switch costs. Aligned signals were concentrated in several large-scale systems, especially subcortical regions, while variability in aligned signals was not associated with switch costs.
- Liberal signals were concentrated largely in subcortical regions, whereas other systems showed distinct alignment or liberality dispositions.Fronto-parietal, cingulo-opercular, and default mode systems tended toward high alignment; visual, auditory, and attention systems showed no such disposition.
- Switch costs were defined as response times during color-cued switch trials versus nonswitching trials.The relationship between aligned and liberal BOLD signals and switch costs was assessed across the brain.
- Variability in aligned signals was not significantly associated with switch costs across subjects.The reported correlation was R = 0.15, p = 0.43, accounting for 2% of the variance.
- R = 0.62 and R = 0.71: lower liberality was associated with lower switch costs during fixation and nonswitching blocks, respectively.These associations were significant for fixation and nonswitching perceptual blocks.
- Aligned signals were concentrated especially in subcortical, default mode, fronto-parietal, and cingulo-opercular systems.The concentration of aligned signals in subcortical structures was confirmed as statistically significant.
- Alignment and liberality were significantly but not perfectly correlated across brain regions, with mean R = 0.69 across subjects and tasks.Subcortical regions contained both highly anatomically aligned and highly liberal signals, indicating multimodal complexity.
derlying cognitive flexibility
Functional alignment with white matter anatomy was associated with faster cognitive switching across individuals. Regional alignment patterns were stable across task conditions, while subcortical and cingulate dynamics showed behaviorally relevant relationships.
- Individuals with more anatomically aligned liberal signals switched perceptual focus faster.Relative alignment with anatomy was associated with greater cognitive flexibility and lower switch costs.
- Behaviorally relevant signals were concentrated in subcortical and cingulate systems, where modest alignment of liberal signals accompanied faster switching.These findings emphasize subcortical functional dynamics atop anatomical network organization.
- Regional patterns of alignment and liberality were consistent across fixation, non-switching, and switching conditions.This consistency was observed within subjects across all three task conditions.
- The anatomically grounded framework distinguishes dynamic contributions of subcortical and other brain systems to cognitive switching.It integrates functional dynamics with white matter structure rather than analyzing either modality alone.
- Highly flexible systems showed strong dependence on underlying anatomical networks across BOLD frames.Moment-to-moment signal configurations in these systems were organized by structure over time.
- The study does not explain the potential cognitive role of highly aligned signals, which may depend on other cognitive control processes.Whether signal liberality generalizes to switching tasks involving other sensory modalities remains to be established.
Methods
The study recruited 28 participants for a local-global perceptual switching task and combined fMRI with diffusion imaging to relate functional signals to anatomical connectivity. Graph signal processing decomposed BOLD data using subject-specific tractography-derived networks.
- Task: Participants performed local-global trials requiring reports of local or global features based on stimulus color, with randomly ordered blocks and switching trials.White stimuli cued local reports, green stimuli cued global reports, and switch blocks changed color across trials with 70% containing a switch.
- Imaging: Diffusion spectrum imaging and fMRI were acquired in the same session, while anatomical and functional images underwent registration, segmentation, smoothing, and artifact-reduction preprocessing.Functional preprocessing included skull stripping, motion correction, slice-timing correction, 6-mm spatial smoothing, and high-pass filtering.
- Network construction: Anatomical connectivity matrices counted streamline connections between parcellated regions, normalized them by regional volumes, and removed cerebellum-to-cerebellum edges.The parcellation included 129 regions and incorporated cerebellar parcels.
- Signal analysis: Graph signal processing analyzed BOLD signals on connected, weighted, symmetric anatomical graphs using eigenvector-based transforms and decomposed them into aligned and liberal components.The graph was defined as G = (V, A), with vertices representing individual brain regions.
Supplementary Information and Analyses
Supplementary analyses characterized behavioral performance, the stability of aligned and liberal signals, and their regional and systems-level variability. These analyses further examined whether signal organization tracked anatomical networks across task conditions.
- Behavioral performance: 94% (St.D. = 1%) accuracy was observed across trials, with median response times of 0.89 s for non-switching and 1.22 s for switching trials.The average switch cost was 0.32 s (St.D. = 0.08 s).
- Signal stability: Aligned and liberal signals formed stable subject-level traits across fixation, no-switch, and switching conditions.The analyses compared signal concentration vectors across regions for each pair of conditions.
- Signal stability: Within each signal type, condition-wise concentration patterns were highly correlated, with mean R = 0.99 for both aligned and liberal signals.The supplied passages state that correlations between the two signal types were only moderate.
- Regional organization: Aligned signals were concentrated in fronto-parietal, cingulo-opercular, default-mode, and subcortical systems and were strongly organized by subject-specific anatomy from TR to TR.These systems included cognitive-control networks described as functionally dynamic.
- Signal variability: Fronto-parietal and cingulo-opercular systems were more variable across TRs than null expectations, whereas ventral-attention, somatosensory, and cerebellar systems were less variable.Variability was assessed using average time-series standard deviations and 10,000 region-to-system permutations.
System flexibility is associated with signal alignment
The study related temporal system flexibility to trial-level alignment between BOLD signals and anatomical networks. Across systems, greater mean alignment accompanied greater mean flexibility.
- The authors tested whether anatomical alignment at the TR level was associated with flexibility expressed over longer temporal scales.
- The analysis constructed subject-specific multilayer temporal networks from correlations across 10 non-overlapping 40-TR windows.Nonsignificant correlations were set to zero after false discovery rate correction at 0.05.
- System flexibility was defined from changes in nodes’ modular assignments across consecutive temporal-network layers.Each system’s flexibility was the mean flexibility of its constituent nodes.
- R = 0.15, p = 0.015 for the positive correlation between mean system flexibility and mean alignment across systems.
Signal alignment in anatomy is similar to function
The study compared signals aligned with anatomical and functional networks to test whether the two network representations organize similar BOLD dynamics. Anatomically and functionally aligned signals were strongly correlated across subjects.
- Graph Fourier analysis links network-aligned signal distributions to eigenvectors associated with large eigenvalues of the correlation matrix.
- The analysis decomposed each subject’s BOLD TRs using both the anatomical network and the mean functional correlation network, then compared the aligned signals.
- R = 0.814, p = 1.613 × 10^-27 for the correlation between anatomically and functionally aligned signals.Signals most aligned with the functional network were also most aligned with the anatomical network.
- The findings indicate that both anatomical and functional network eigenspectra organize flexible activity in the human brain.
Aligned and liberal signal associations with behavior across task
The study examined associations between graph-derived signal components and response-time behavior across 28 subjects and task conditions. Liberal signals, rather than aligned signals, showed the reported relationship with switch-cost variability.
- The behavioral analyses associated median response times in no-switch, switching, and switch-cost conditions with aligned and liberal signals.
- Only liberal signals demonstrated a relationship with behavioral variability, specifically switch costs, within the anatomical signal decomposition.
- The reported partial correlations used average framewise displacement as a covariate and were assessed across signal types and task conditions.
- Results remained significant with similar correlation values after additionally controlling for age and sex.
Robustness of aligned and liberal signals to parameter selection
The authors tested whether signal decomposition and anatomical alignment findings were robust to parameter choices and network randomization. Liberal and aligned signals were stable, and real anatomical networks explained more aligned-signal variance than null networks.
- The observed signal decomposition was tested across choices of KL and KH ranging from five below to five above the main-manuscript parameters.
- Variance accounted for by aligned-signal eigenvectors in real anatomical networks exceeded that in 100% of degree- and strength-preserving null permutations.
- The result indicates that anatomically organized BOLD signals reflect contributions from network topology beyond direct node connections.
- The authors state that the behavioral relevance of anatomically aligned organization requires establishment in future studies.
Alternate measures of function and anatomy
Simple measures of BOLD variation, local anatomical influence, and their relationship did not account for switch-cost variability, whereas whole-network signal decomposition provided greater associative value.
- Mean BOLD variance in subcortical regions showed no significant relationship with switch costs (R = −0.09, p = 0.68).
- 0.547% of variance was accounted for by aligned components in true anatomical networks, compared with 0.455% under randomized networks.
- Mean subcortical node strength was not significantly related to switch costs (R = −0.15, p = 0.48).Node strength was defined as the sum of connections to each node.
- The relationship between BOLD signals and anatomical node degree across subcortical regions showed no significant association with switch costs (R = 0.17, p = 0.41).
- Together, simple BOLD, local anatomy, and pairwise BOLD–anatomy measures were insufficient to explain switch-cost variability, while whole-network decomposition had superior associative value.
switch costs.
A degree- and strength-preserving network permutation test assessed whether the observed behavioral association depended on the specific anatomical configuration. The observed correlation exceeded 99% of null correlations, and remained similar after accounting for motion.
- 200 randomized anatomical networks preserved each subject’s original strength and degree distributions before signal decomposition and behavioral correlation.
- The observed correlation between liberal signals and switch costs exceeded 99% of correlations from the null distribution.
- The null test indicated that the specific configuration of the true anatomical network drives the correlation between liberal signal alignment and switch costs.
- The observed correlation was R = 0.57, p = 0.002 with motion covariate and R = 0.59, p = 0.001 without it.Motion correlated significantly with liberal signals (R = 0.77, p = 1.10×10−6) but not significantly with switch costs (R = 0.26, p = 0.16).
- The behaviorally relevant portion of liberal signals was not driven by motion.