Network models are a key tool in human neuroscience, and translation into animal models is essential for interrogating mechanistic drivers of network organization. Using magnetic resonance imaging (MRI), we present the first in vivo network representation of the individual rat brain, a key animal model in neuroscience. We measured magnetization transfer ratio (MTR) at each of 53 distinct cortical areas and estimated a cortical similarity network for each scan across two independent cohorts. We characterized normative network development in rats scanned repeatedly between postnatal days 20 (weanling) and 290 (mid-adulthood; N = 47) and then contrasted these findings with a cohort exposed to early life stress (repeated maternal separation [RMS]; N = 40). The normative rat cortical similarity network exhibited biologically meaningful organization, consistent with cytoarchitectonic and tract-tracing data, and displayed complex topological features. Developmental analyses revealed increasing interregional similarity during early postnatal and adolescent periods, followed by divergence in mid-adulthood, particularly within fronto-hippocampal systems. RMS disrupted these trajectories, especially between frontal and parahippocampal regions that were also most dynamic during development and aging. These findings introduce a new network-based methodology for studying cortical organization in a model organism, providing a translational framework to understand how environmental risk factors alter brain network development.
The thalamus is a heterogeneous structure crucial for corticocortical communication, affective-perceptual integration, motor preparation, and memory-related functions. Posttraumatic stress disorder (PTSD) is characterized by various symptoms that likely relate to thalamic functions. Group-level and individual differential structural covariance (SC) analyses were conducted on intrathalamic, thalamocortical, and thalamosubcortical volumetric networks by segmenting structural MRI data from 2,784 subjects (PTSD n = 1,306; controls n = 1,478) into 25 thalamic nuclei per hemisphere. We found that PTSD was associated with stronger intrathalamic and thalamocortical network strength and stronger SC between the limbic thalamus and the somatomotor and auditory thalamus. PTSD severity was related to specific regional alterations in the intrathalamic network involving the lateral pulvinar. Comorbid depression severity positively correlated with global intrathalamic alterations, while avoidance symptoms positively correlated with global thalamosubcortical alterations. Hyperarousal symptoms related to altered SC in the thalamocortical network between the reuniens, central medial, paratenial, centromedian, and limitans-suprageniculate nuclei and lateral cortical regions spanning the occipital, temporal, and orbitofrontal cortices. Differential associations between avoidance, hyperarousal, and comorbid depression symptoms and thalamic SC in PTSD suggest that specific thalamic covariance patterns may be involved in unique facets of PTSD symptomatology.
Alzheimer’s disease (AD) is characterized by progressive neural network degradation. In brain functional networks, overlapping module structures provide more accurate representations of brain function than nonoverlapping structures. Since the involvement of overlapping nodes in multiple modules can vary over time, investigating dynamic functional changes in the brain may provide deeper insights into the structural characteristics of these overlapping modules. However, the spatiotemporal dynamics of overlapping modular brain organization remain unclear. We employed resting-state fMRI to explore the overlapping modular organization and dynamic multilayer modules in 64 AD (Agemean = 74.04) and 61 healthy controls (HC, Agemean = 74.86) from the Alzheimer’s Disease Neuroimaging Initiative. Compared with HC, AD exhibited increased overlapping modules and decreased modularity, with altered nodal overlapping probability, particularly in the superior frontal cortex and hippocampus. Higher nodal overlapping probability correlated with greater flexibility and was associated with larger amyloid deposits. Lasso regression analysis further revealed strong correlations between overlapping nodal characteristics and cognitive performance. Our findings suggest that overlapping nodes are critical components in AD, demonstrating high amyloid deposition, significant functional flexibility, and strong associations to cognitive behavior. These alterations may enhance the understanding of AD pathology and contribute to the development of biomarkers for improved diagnosis and therapeutic strategies.
When using a cortical parcellation, it is generally assumed that the regions correspond across subjects, meaning regions with the same label are expected to have the same structural or functional roles from one subject to any other. However, at a fine-grained scale, environmental and genetic factors shape both cortex and white matter differently, making it challenging to consistently label all parcels for every subject. Brain networks can be constructed with regions as nodes and their connectivity as edges. One critical step while constructing those networks is the definition of the nodes due to the spatial inter-individual variability, which can limit the reliability of group studies’ results. In this work, we propose alignment criteria to register the brain nodes across subjects by allowing the permutation of tiny cortical regions. Those criteria account for multiple brain network perspectives built from different imaging modalities. Our across-subject multimodal alignment of brain networks includes constraints that restrict possible permutations to anatomically plausible ones. The identified permutations are finally applied to the brain networks not used for optimization, and also improve the alignment of these networks. This work has been validated on real magnetic resonance imaging data from the Human Connectome Project.
The interactive effects of HIV and marijuana (MJ) on the brain remain largely unknown, despite the prevalence of cognitive implications in this population. This study examined the impacts of HIV and MJ on brain networks crucial for normal cognitive function. Functional MRI data and a battery of neuropsychological tests from 237 HIV+ and HIV− adults aged 25–59 years, stratified by MJ use, were collected. The following hypotheses were then tested: (a) HIV is associated with widespread disruption of the small-world organization of the default mode network (DMN) that is exacerbated by MJ use; (b) observed differences are reflected in cognitive performance. Clustering coefficient and global efficiency were used to measure small-world organizations. We found significant differences in the DMN’s clustering and efficiency between our control group (HIV−MJ−) and the other three groups (HIV+MJ−, HIV−MJ+, and HIV+MJ+). In those with HIV/MJ, the DMN reorganized toward a network explained by efficiency than clustering. Global cognitive performance was associated with this group difference. Unlike the control group, participants with HIV/MJ showed an integrated DMN across all cognitive scores. The higher integrity of the DMN in patients with HIV and MJ use (particularly when co-occurring) across cognitive scores could imply compensation to preserve cognitive function.
Estimation of brain connectivity from neuroelectric data is a fundamental problem in modern neuroscience, and it is used to assess the network properties of brain function. In the present work, we critically assess the virtues and limitations of temporal Granger causality (using both conditional and unconditional formulations) for the estimation of functional brain connectivity, using a neural mass model as the ground truth. The model simulates transmission among different brain rhythms (in the θ, α, β, and γ bands) via excitatory and inhibitory synapses. The results show that Granger causality is able to detect relative changes in connectivity, but the estimated values are influenced by the operative conditions (sampling frequency, signal length, delay). Moreover, the absolute value of Granger causality depends on the particular rhythm transmitted and is affected by nonlinear phenomena, especially the activity level in the connected regions. In the case of complex connectivity networks, conditional Granger causality overwhelms the unconditional one, since the latter often discovers spurious connections. Finally, inhibitory connections can be revealed more easily by Granger causality than similar excitatory connections, a result generally neglected in brain network studies. The present results can drive the correct interpretation of Granger-causality-based connectivity networks derived from neuroelectric signals.
Overcoming an initial psychotic episode does not always lead to recovery; relapses and subsequent psychotic episodes may happen afterward. Even if the characterization of psychotic disorders can be related to alterations in brain connectivity, clear identification of the brain areas for relapse is missing. Here, we leverage on whole-brain modeling linking anatomical structural information with functional activity as measured by MRI in 196 participants. Patients were classified into Stage II (first episode), IIIa (incomplete remission), IIIb (remission followed by one relapse), and IIIc (remission followed by several relapses), depending on the course of psychosis up to the time of the brain scan. From these data, a low-dimensional manifold reduction of the brain dynamics was obtained using deep learning variational autoencoders in which the different stages are represented, and a classification model can be trained to distinguish them. Then, a dimensionality analysis was performed to find the optimal dimension that allows the distinction between first episode and relapsing cases with high accuracy. Finally, perturbations were introduced in the model to reveal the brain regions associated with the absence of relapse, which could help predict which brain regions to target during therapy and assist the treatment of patients suffering from psychotic disorders.
Numerous methods exist for quantifying statistical dependencies, termed "functional coupling" (FC), between regional brain activity recorded with resting-state functional magnetic resonance imaging (rs-fMRI). However, their efficacy in mitigating the effects of known sources of noise, such as those induced by participant head motion, and in augmenting effect sizes for brain-wide association studies (BWAS), remains unclear. Here we compared 10 different measures of FC, including correlations, partial correlations, coherence, mutual information, and partial information decomposition, and one measure of effective connectivity (EC; regression dynamic causal modeling), across two independent datasets comprising a total of 1,797 participants (867 males). Each method was evaluated for its ability to mitigate motion-related confounds in FC/EC estimates (assessed via framewise-displacement - edgewise FC correlations) and for its utility in predicting 94 behavioral measures (assessed via cross-validated kernel ridge regression). Our analyses showed that EC was most resistant to motion artifacts but had the weakest behavioral predictions. Conversely, traditional correlation-based methods showed the highest sensitivity to motion, but offered the strongest behavioral prediction across most domains and datasets. Nonetheless, relative differences in predictive accuracies were small, indicating that the use of different FC or EC metrics in rs-fMRI does not significantly impact BWAS effect sizes.
Decades of neuroimaging have revealed that the functional organization of the brain is roughly consistent across individuals, and at rest, it resembles group-level task-evoked networks. A fundamental assumption in the field is that the functional specialization of a brain region arises from its connections to the rest of the brain, but limitations in the amount of data that can be feasibly collected in a single individual leave open the following question: Is the association between task activation and connectivity consistent across the brain and many cognitive tasks? To answer this question, we fit ridge regression models to activation maps from 33 cognitive domains (generated with NeuroQuery) using resting-state functional connectivity data from the Human Connectome Project as the predictor. We examine how well functional connectivity fits activation and find that all regions and all cognitive domains have a very robust relationship between brain activity and connectivity. The tightest relationship exists for higher order, domain-general cognitive functions. These results support the claim that connectivity is a general organizational principle of brain function by comprehensively testing this relationship in a large sample of individuals for a broad range of cognitive domains and provide a reference for future studies engaging in individualized predictive models.
Brain function is inherently dynamic, characterized by transient, overlapping functional states rather than static connectivity patterns. Current clustering-based dynamic functional network connectivity methods often fail to capture overlapping states; meanwhile, independent component analysis (ICA)-based methods typically rely on group-level analysis, limiting subject-specific accuracy. To address this gap, we introduce a novel analytical framework estimating individualized dynamic double functional independent primitives (ddFIPs)-based states. Our methodological innovation includes: (a) a two-stage ICA combining spatially constrained ICA to define group-level intrinsic connectivity networks (ICNs), followed by constrained ICA to estimate subject-specific states and timecourses; (b) calibration ensuring derived states preserve original correlation scales, enabling meaningful cross-subject and group-level comparisons; and (c) novel metrics leveraging this calibrated representation, including amplitude convergence (uniformity of simultaneous state contributions), amplitude divergence (variability of states independent of state dominance), and dynamic state density (number of concurrently active states at any given time). These methodological advances enhance our ability to characterize subtle differences in brain connectivity dynamics, offering deeper insight into healthy and disrupted connectivity patterns. Validating our framework on an extensive resting-state fMRI dataset (N > 5.5K) spanning four neuropsychiatric conditions revealed disorder-specific connectivity signatures: schizophrenia exhibited extensive variability (increased divergence), while autism displayed pronounced stability (increased convergence). In summary, our proposed method uniquely integrates subject-specific ICA estimation, unit-preserving calibration, and novel convergence-divergence metrics, providing data-driven biomarkers differentiating psychiatric disorders.
Brain networks exhibit non-trivial modular organization, with groups of densely connected areas participating in specialized functions. Traditional community detection algorithms assign each node to one module, but this representation cannot capture integrative, multifunctional nodes that span multiple communities. Despite the increasing availability of overlapping community detection algorithms (OCDAs) to capture such integrative nodes, there is no objective procedure for selecting the most appropriate method and its parameters for a given problem. Here, we overcome this limitation by introducing a data-driven method for selecting an OCDA and its parameters from performance on a tailored ensemble of generated benchmark networks, assessing 22 unique algorithms and parameter settings. Applied to the human right-hemisphere structural connectome, we find that the "order statistics local optimization method" (OSLOM) best identifies ground-truth overlapping structure in the benchmark ensemble, yielding a seven-network decomposition of the right-hemisphere cortex. These modules are bridged by 15 overlapping regions that generally sit at the apex of the putative cortical hierarchy-suggesting integrative, higher order function-with network participation increasing along the cortical hierarchy, a finding not supported using a non-overlapping modular decomposition. This data-driven approach to selecting OCDAs is applicable across domains, opening new avenues to detecting and quantifying informative structures in complex real-world networks.
The creation and curation of synaptic-level neuronal networks, or connectomes, enables the study of the relationship between neuronal structure and function. Topological characteristics of neuronal networks have been studied extensively. Separately, there have been considerable efforts to classify the morphology, cell types, and lineages of neurons. Here, we introduce a network metric that combines topological analysis with node metadata. This entropic quantity measures the diversity of incoming or outgoing connections to a node in terms of the metadata distribution. We find that in Caenorhabditis elegans, the top-scoring neurons (PVR, RMGL/R, DVA, CEP, ADE, URXR, RIGL, BAG, SMBDL) have known functions that integrate and disseminate multimodal information involved in sensorimotor functions. In the nerve cord of Drosophila melanogaster, we find that top-scoring neurons are embryonic neurons located in the abdominal neuropil, where sensorimotor coordination is required for complex innate behavior such as mating.
Edge time series decompose functional connections into their fine-scale, framewise contributions. Previous studies have demonstrated that global high-amplitude “events” in edge time series can be clustered into distinct patterns. However, whether events and their patterns change or persist throughout the human lifespan has not been investigated. Here, we directly address this question by clustering event frames using the Nathan Kline Institute-Rockland sample that includes subjects with ages spanning the human lifespan. We find evidence of two main clusters that appear across subjects and age groups which systematically change in magnitude and frequency with age. Our results also demonstrate that such event clusters have distinct, heterogeneous relationships with structural connectivity-derived communication measures, which change with age. Finally, event clusters were found to outperform nonevents in predicting phenotypes regarding human intelligence and achievement. Collectively, our findings fill several gaps in current knowledge about cofluctuation patterns in edge time series and human aging, setting the stage for future investigation into the causal origins of changes in functional connectivity throughout the human lifespan.
Network control theory (NCT) models human connectomes as high-dimensional input-state-output stable systems where the efficiency of neural connections can be addressed by energy cost (of state transitions) and controllability (from/to reachable states). Different options are available to extract NCT features: initial/final states, control time horizon, structural (vs. functional), and static (vs. dynamic) connectivity measure. Leveraging the minimum control paradigm, assuming the Schur stability for discrete systems, we investigate intra- and inter-individual variability of NCT features, across different settings and datasets, and assess their potential as useful connectome metrics in clinical studies. NCT was applied to structural and functional MRI (fMRI), in a cohort of 82 cognitively unimpaired elderly subjects with normal or (age-related) sensorineural condition (hearing loss), and in young adults from the Human Connectome Project database. Results demonstrated low intra-individual and moderate within-group inter-individual variability of NCT features. The energy cost was related to the time horizon of the system but did not discriminate groups. Controllability analyses revealed significant group effects and acceptable discrimination between normal and disease-affected connectomes, particularly for the default-mode network. We provide a systematic evaluation of different settings for fMRI-derived NCT features that may help guiding clinical applications toward capturing neurologically meaningful changes in the human connectome.
Despite considerable efforts to uncover the neural basis of psychiatric disorders using neuroimaging, few methods utilize intrinsic brain-derived knowledge, leading to limited specificity and discriminability in biomarker identification. To leverage the inherent characteristics within the brain, we propose a prior-knowledge-guided feature selection method to flexibly unveil discriminative and target-oriented biomarkers of psychiatric disorders. Specifically, we construct a constrained sparse regularization allowing for the flexible integration of diverse prior knowledge to identify sparse neuroimaging features linked to specific psychopathology. Additionally, we simultaneously integrate graph-based regularization and redundancy-removal regularization to further ensure the discriminability and independence among the selected features. Different priors hold varying significance in identifying specific biomarkers. Four functional magnetic resonance imaging (fMRI) datasets from 708 healthy controls and 537 schizophrenia patients are used to evaluate our method integrated with various prior knowledge, revealing specific schizophrenia-related brain abnormalities. Compared with nine advanced feature selection methods, our method improves mean classification accuracy by 3.89% to 11.24%, particularly revealing reduced interactions within the visual domain and between subcortical and visual domains in schizophrenia patients. The proposed method offers flexible and precise biomarker identification tailored to specific targets, advancing the understanding and diagnosis of psychiatric conditions.
The brain’s ability to undergo complex rewiring during development is a fascinating aspect of neuroscience. This study conducts a detailed comparison of Drosophila melanogaster’s brain networks during larval and adult stages, revealing significant changes in neuronal wiring throughout development. The larval brain network exhibits a degree distribution that fits firmly to a Weibull model. In contrast, the sparser adult brain network follows a power-law distribution, with the out-degree exponent lying in the scale-free regime and the in-degree exponent close to it. This shift toward a scale-free pattern likely reflects an adaptation to enhance robustness against failures while minimizing costs associated with reduced density during development. We also observed alterations in the structural core in relation to cell composition and topological influence. The structural core of the larva comprises neurons in the mushroom body, while neurons in the antennal lobe form the core of the adult fly brain.Furthermore, the larval network solely shows a rich club organization of which the structural core is also a part. Analysis of connectivity, rich club, and network measures reveals that the shift in the core results from a reduction in the centrality of mushroom body neurons following metamorphosis. This work stands as a step forward in understanding the rewiring of brain networks across the life stages of D. melanogaster.
Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by deficits in social cognition, interaction, communication, restricted behaviors, and sensory abnormalities. The heterogeneity in ASD's clinical presentation complicates its diagnosis and treatment. Recent technological advancements in graph neural networks (GNNs) have been extensively used to diagnose brain disorders such as ASD, but existing machine learning models often suffer from low accuracy and explainability. In this study, we proposed a novel, explainable, and generalized node-edge connectivity-based graph attention neural network (Ex-NEGAT) model, leveraging edge-centric high-order spatiotemporal organization of dynamic functional connectivity streams between large-scale functional brain networks implicated in autism. Using the Autism Brain Imaging Data Exchange I and II datasets (total samples = 1,500), the model achieved 88% accuracy and an F1-score of 0.89. Additionally, we used meta-connectivity subtypes to identify subgroups within ASD samples using the rough fuzzy c-means algorithm. We also used connectome-based prediction modeling, which revealed critical brain networks contributing to predictions that accurately correlate with Autism Diagnostic Observation Schedule (ADOS) and full intelligent quotient (FIQ) scores. The proposed framework offers a robust approach based on previously unexplored higher order spatiotemporal correlation features of dynamic functional connectivity, which may provide critical insight into ASD heterogeneity and improve diagnostic precision.
Brain connectivity analysis is pivotal to understanding mechanisms underpinning neurological diseases. However, current methodologies for quantitatively mapping the connectivity in vivo face challenges when focal lesions are present and can introduce strong biases in the estimates. We present a novel approach to address these challenges by introducing a multi-compartment description of the connectome, which explicitly incorporates lesion information during the estimation process. We extended the Convex Optimization Modeling for Microstructure Informed Tractography (COMMIT) framework to integrate an additional tissue compartment in voxels affected by pathology, allowing us to infer accurately the contributions of streamlines passing through lesions and to provide unbiased connectivity estimates. We evaluated the effectiveness of our approach on data from healthy subjects of the Human Connectome Project, in which we artificially introduced focal lesions to simulate pathology with varying levels of axonal damage. We also tested the performances obtained when comparing healthy subjects with patients affected by multiple sclerosis. Results demonstrate that our approach significantly enhances sensitivity to pathological changes even at low degeneracy levels compared with state-of-the-art techniques, thus representing a significant step forward to advance our understanding of neurodegenerative diseases.
Reading difficulty (RD; dyslexia) is a developmental condition with neurological origins and persistent academic consequences. Children with RD often show deficits in audiovisual integration (AVI) and executive functions. Visual rhythmic reading training (RRT) has been associated with improvements in these domains, but it remains unclear whether such effects generalize to the resting-state brain activity. English-speaking children aged 8-12 years, including typical readers (TRs) and children with RD, were randomly assigned to an 8-week visual RRT or control math training group. Reading assessments and resting-state functional MRI data were collected before and after the intervention. Functional connectivity (FC) analyses examined AVI and its interaction with frontoparietal-cingulo-opercular (FP-CO) cognitive control networks during rest. Following RRT, children with RD showed significant improvements in reading fluency. The RRT group also demonstrated greater changes in AVI, which were associated with increased FC between FP-CO networks and sensory regions during the resting state. RRT improves reading performance and promotes enhanced integration between sensory and executive networks in children with RD, even in the absence of task demands. These findings support the role of RRT in fostering domain-general neuroplasticity beyond reading-specific contexts.
