Classification of Brain Disorders in rs-fMRI via Local-to-Global Graph Neural Networks

IEEE Transactions on Medical Imaging, Journal Year: 2022, Volume and Issue: 42(2), P. 444 - 455

Published: Nov. 4, 2022

Recently, functional brain network has been used for the classification of disorders, such as Autism Spectrum Disorder (ASD) and Alzheimer's disease (AD). Existing methods either ignore non-imaging information associated with subjects relationship between subjects, or cannot identify analyze disease-related local regions biomarkers, leading to inaccurate results. This paper proposes a local-to-global graph neural (LG-GNN) address this issue. A ROI-GNN is designed learn feature embeddings global Subject-GNN then established generated by information. The contains self-attention based pooling module preserve most important classification. an adaptive weight aggregation block generate multi-scale embedding corresponding each subject. proposed LG-GNN thoroughly validated using two public datasets ASD AD experimental results demonstrated that it achieves state-of-the-art performance in terms various evaluation metrics.

Language: Английский

---

Information decomposition and the informational architecture of the brain

Trends in Cognitive Sciences, Journal Year: 2024, Volume and Issue: 28(4), P. 352 - 368

Published: Jan. 9, 2024

To explain how the brain orchestrates information-processing for cognition, we must understand information itself. Importantly, is not a monolithic entity. Information decomposition techniques provide way to split into its constituent elements: unique, redundant, and synergistic information. We review disentangling redundant interactions redefining our understanding of integrative function neural organisation. navigates trade-offs between redundancy synergy, converging evidence integrating structural, molecular, functional underpinnings synergy redundancy; their roles in cognition computation; they might arise over evolution development. Overall, provides guiding principle informational architecture cognition.

Language: Английский

---

The architecture of the human default mode network explored through cytoarchitecture, wiring and signal flow

Nature Neuroscience, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 28, 2025

Abstract The default mode network (DMN) is implicated in many aspects of complex thought and behavior. Here, we leverage postmortem histology vivo neuroimaging to characterize the anatomy DMN better understand its role information processing cortical communication. Our results show that cytoarchitecturally heterogenous, containing cytoarchitectural types are variably specialized for unimodal, heteromodal memory-related processing. Studying diffusion-based structural connectivity combination with cytoarchitecture, found contains regions receptive input from sensory cortex a core relatively insulated environmental input. Finally, analysis signal flow effective models showed unique amongst networks balancing output across levels hierarchies. Together, our study establishes an anatomical foundation which accounts broad plays human brain function cognition can be developed.

Language: Английский

---

Local vulnerability and global connectivity jointly shape neurodegenerative disease propagation

PLoS Biology, Journal Year: 2019, Volume and Issue: 17(11), P. e3000495 - e3000495

Published: Nov. 21, 2019

It is becoming increasingly clear that brain network organization shapes the course and expression of neurodegenerative diseases. Parkinson disease (PD) marked by progressive spread atrophy from midbrain to subcortical structures and, eventually, cerebral cortex. Recent discoveries suggest process involves misfolding prion-like propagation endogenous α-synuclein via axonal projections. However, mechanisms translate local "synucleinopathy" large-scale dysfunction remain unknown. Here, we use an agent-based epidemic spreading model integrate structural connectivity, functional gene predict sequential volume loss due neurodegeneration. The dynamic replicates spatial temporal patterning empirical in PD implicates substantia nigra as epicenter. We reveal a significant role for both connectome topology geometry shaping distribution atrophy. also demonstrates SNCA GBA transcription influence concentration regional vulnerability. Functional coactivation further amplifies set architecture expression. Altogether, these results support theory progression multifactorial depends on cell-to-cell misfolded proteins

Language: Английский

---

Structure-function coupling in the human connectome: A machine learning approach

NeuroImage, Journal Year: 2020, Volume and Issue: 226, P. 117609 - 117609

Published: Dec. 1, 2020

While the function of most biological systems is tightly constrained by their structure, current evidence suggests that coupling between structure and brain networks relatively modest. We aimed to investigate whether modest connectome a fundamental property nervous or limitation network models. developed new deep learning framework predict an individual's from structural connectome, achieving prediction accuracies substantially exceeded state-of-the-art biophysical models (group: R=0.9±0.1, individual: R=0.55±0.1). Crucially, predicted explained significant inter-individual variation in cognitive performance. Our results suggest structure-function human tighter than previously suggested. establish margin which can be improved demonstrate how facilitate investigation relations behavior.

Language: Английский

---

Inferring neural signalling directionality from undirected structural connectomes

Nature Communications, Journal Year: 2019, Volume and Issue: 10(1)

Published: Sept. 19, 2019

Abstract Neural information flow is inherently directional. To date, investigation of directional communication in the human structural connectome has been precluded by inability non-invasive neuroimaging methods to resolve axonal directionality. Here, we demonstrate that decentralized measures network communication, applied undirected topology and geometry brain networks, can infer putative directions large-scale neural signalling. We propose concept send-receive asymmetry characterize cortical regions as senders, receivers or neutral, based on differences between their incoming outgoing efficiencies. Our results reveal a hierarchy recapitulates established organizational gradients differentiating sensory-motor multimodal areas. find asymmetries are significantly associated with directionality effective connectivity derived from spectral dynamic causal modeling. Finally, using fruit fly, mouse macaque connectomes, provide further evidence suggesting signalling encoded architecture nervous systems.

Language: Английский

---

Network communication models improve the behavioral and functional predictive utility of the human structural connectome

Network Neuroscience, Journal Year: 2020, Volume and Issue: 4(4), P. 980 - 1006

Published: Jan. 1, 2020

The connectome provides the structural substrate facilitating communication between brain regions. We aimed to establish whether accounting for polysynaptic in connectomes would improve prediction of interindividual variation behavior as well increase structure-function coupling strength. Connectomes were mapped 889 healthy adults participating Human Connectome Project. To account signaling, transformed into matrices each 15 different network models. Communication (a) used perform predictions five data-driven behavioral dimensions and (b) correlated resting-state functional connectivity (FC). While FC was most accurate predictor behavior, models, particular communicability navigation, improved performance connectomes. also strengthened coupling, with navigation shortest paths models leading 35–65% increases association strength FC. combined results a single ranking that insight which may more faithfully recapitulate underlying neural signaling patterns. Comparing across multiple mapping pipelines suggested modeling is particularly beneficial sparse high-resolution conclude can augment predictive utility human connectome.

Language: Английский

---

A dynamic graph convolutional neural network framework reveals new insights into connectome dysfunctions in ADHD

NeuroImage, Journal Year: 2021, Volume and Issue: 246, P. 118774 - 118774

Published: Nov. 30, 2021

The pathological mechanism of attention deficit hyperactivity disorder (ADHD) is incompletely specified, which leads to difficulty in precise diagnosis. Functional magnetic resonance imaging (fMRI) has emerged as a common neuroimaging technique for studying the brain functional connectome. Most existing methods that have either ignored or simply utilized graph structure, do not fully leverage potentially important topological information may be useful characterizing disorders. There crucial need designing novel and efficient approaches can capture such information. To this end, we propose new dynamic convolutional network (dGCN), trained with sparse regional connections from dynamically calculated features. We also develop readout layer improve representation. Our extensive experimental analysis demonstrates significantly improved performance dGCN ADHD diagnosis compared machine learning deep methods. Visualizations salient regions interest (ROIs) connectivity based on informative features learned by our model show identified abnormalities mainly involve temporal pole, gyrus rectus, cerebellar gyri lobe, frontal cerebellum, respectively. A positive correlation was further observed between connectomic symptom severity. proposed shows great promise providing network-based precision broadly applicable connectome-based study mental

Language: Английский

---

Network science characteristics of brain-derived neuronal cultures deciphered from quantitative phase imaging data

Scientific Reports, Journal Year: 2020, Volume and Issue: 10(1)

Published: Sept. 15, 2020

Abstract Understanding the mechanisms by which neurons create or suppress connections to enable communication in brain-derived neuronal cultures can inform how learning, cognition and creative behavior emerge. While prior studies have shown that possess self-organizing criticality properties, we further demonstrate vitro exhibit a self-optimization phenomenon. More precisely, analyze multiscale neural growth data obtained from label-free quantitative microscopic imaging experiments reconstruct culture networks (microscale) cluster (mesoscale). We investigate structure evolution of estimating importance each network node their information flow. By analyzing degree-, closeness-, betweenness-centrality, node-to-node degree distribution (informing on interconnection phenomena), clustering coefficient/transitivity (assessing “small-world” properties), multifractal spectrum, murine self-optimizing over time with topological characteristics distinct existing complex models. The time-evolving among optimizes flow, robustness, self-organization degree. These findings implications for modeling potentially design biological inspired artificial intelligence.

Language: Английский

---

Signal diffusion along connectome gradients and inter-hub routing differentially contribute to dynamic human brain function

NeuroImage, Journal Year: 2020, Volume and Issue: 224, P. 117429 - 117429

Published: Oct. 7, 2020

Human cognition is dynamic, alternating over time between externally-focused states and more abstract, often self-generated, patterns of thought. Although cognitive neuroscience has documented how networks anchor particular modes brain function, mechanisms that describe transitions distinct functional remain poorly understood. Here, we examined time-varying changes in function emerge within the constraints imposed by macroscale structural network organization. Studying a large cohort healthy adults (n = 326), capitalized on manifold learning techniques identify low dimensional representations connectome organization decomposed neurophysiological activity into their transition using Hidden Markov Models. Structural predicted dynamic anchored sensorimotor systems those transmodal states. Connectome topology analyses revealed involving traversed short intermediary distances adhered strongly to communication diffusion. Conversely, involved spatially distributed hubs increasingly engaged long-range routing. These findings establish structure cortex optimized allow neural freedom vary processing, so provides key insight give rise flexibility human cognition.

Language: Английский

---