A graph neural network framework for causal inference in brain networks

AbstractA central question in neuroscience is how self-organizing dynamic interactions in the brain emerge on their relatively static structural backbone. Due to the complexity of spatial and temporal dependencies between different brain areas, fully comprehending the interplay between structure and function is still challenging and an area of intense research. In this paper we present a graph neural network (GNN) framework, to describe functional interactions based on the structural anatomical layout. A GNN allows us to process graph-structured spatio-temporal signals, providing a possibility to combine structural information derived from diffusion tensor imaging (DTI) with temporal neural activity profiles, like that observed in functional magnetic resonance imaging (fMRI). Moreover, dynamic interactions between different brain regions discovered by this data-driven approach can provide a multi-modal measure of causal connectivity strength. We assess the proposed model’s accuracy by evaluating its capabilities to replicate empirically observed neural activation profiles, and compare the performance to those of a vector auto regression (VAR), like that typically used in Granger causality. We show that GNNs are able to capture long-term dependencies in data and also computationally scale up to the analysis of large-scale networks. Finally we confirm that features learned by a GNN can generalize across MRI scanner types and acquisition protocols, by demonstrating that the performance on small datasets can be improved by pre-training the GNN on data from an earlier study. We conclude that the proposed multi-modal GNN framework can provide a novel perspective on the structure-function relationship in the brain. Accordingly this approach appears to be promising for the characterization of the information flow in brain networks.

Download Full-text

Hemispheric Asymmetry of Human Brain Anatomical Network Revealed by Diffusion Tensor Tractography

BioMed Research International ◽

10.1155/2015/908917 ◽

2015 ◽

Vol 2015 ◽

pp. 1-11 ◽

Cited By ~ 7

Author(s):

Ni Shu ◽

Yaou Liu ◽

Yunyun Duan ◽

Kuncheng Li

Keyword(s):

Human Brain ◽

Large Scale ◽

Diffusion Tensor ◽

Functional Integration ◽

Brain Regions ◽

Hemispheric Asymmetries ◽

Characteristic Path Length ◽

Topological Measures ◽

Structural Connections ◽

The Brain

The topological architecture of the cerebral anatomical network reflects the structural organization of the human brain. Recently, topological measures based on graph theory have provided new approaches for quantifying large-scale anatomical networks. However, few studies have investigated the hemispheric asymmetries of the human brain from the perspective of the network model, and little is known about the asymmetries of the connection patterns of brain regions, which may reflect the functional integration and interaction between different regions. Here, we utilized diffusion tensor imaging to construct binary anatomical networks for 72 right-handed healthy adult subjects. We established the existence of structural connections between any pair of the 90 cortical and subcortical regions using deterministic tractography. To investigate the hemispheric asymmetries of the brain, statistical analyses were performed to reveal the brain regions with significant differences between bilateral topological properties, such as degree of connectivity, characteristic path length, and betweenness centrality. Furthermore, local structural connections were also investigated to examine the local asymmetries of some specific white matter tracts. From the perspective of both the global and local connection patterns, we identified the brain regions with hemispheric asymmetries. Combined with the previous studies, we suggested that the topological asymmetries in the anatomical network may reflect the functional lateralization of the human brain.

Download Full-text

Towards an information theoretical description of communication in brain networks

Network Neuroscience ◽

10.1162/netn_a_00185 ◽

2021 ◽

pp. 1-23

Author(s):

Enrico Amico ◽

Kausar Abbas ◽

Duy Anh Duong-Tran ◽

Uttara Tipnis ◽

Meenusree Rajapandian ◽

...

Keyword(s):

Information Transfer ◽

Large Scale ◽

Brain Networks ◽

Brain Regions ◽

Brain Network ◽

Theoretical Description ◽

Structural Topology ◽

Communication Dynamics ◽

The Brain ◽

The Way

Modeling communication dynamics in the brain is a key challenge in network neuroscience. We present here a framework that combines two measurements for any system where different communication processes are taking place on top of a fixed structural topology: Path Processing Score (PPS) estimates how much the brain signal has changed or has been transformed between any two brain regions (source and target); Path Broadcasting Strength (PBS) estimates the propagation of the signal through edges adjacent to the path being assessed. We use PPS and PBS to explore communication dynamics in large-scale brain networks. We show that brain communication dynamics can be divided into three main “communication regimes” of information transfer: absent communication (no communication happening); relay communication (information is being transferred almost intact); transducted communication (the information is being transformed). We use PBS to categorize brain regions based on the way they broadcast information. Subcortical regions are mainly direct broadcasters to multiple receivers; Temporal and frontal nodes mainly operate as broadcast relay brain stations; Visual and somato-motor cortices act as multi-channel transducted broadcasters. This work paves the way towards the field of brain network information theory by providing a principled methodology to explore communication dynamics in large-scale brain networks.

Download Full-text

On the structural connectivity of large-scale models of brain networks at cellular level

Scientific Reports ◽

10.1038/s41598-021-83759-z ◽

2021 ◽

Vol 11 (1) ◽

Author(s):

Giuseppe Giacopelli ◽

Domenico Tegolo ◽

Emiliano Spera ◽

Michele Migliore

Keyword(s):

Large Scale ◽

Brain Networks ◽

Structural Connectivity ◽

Network Models ◽

Brain Regions ◽

Cellular Level ◽

Scale Model ◽

Mathematical Framework ◽

Underlying Mechanisms ◽

Experimental Findings

AbstractThe brain’s structural connectivity plays a fundamental role in determining how neuron networks generate, process, and transfer information within and between brain regions. The underlying mechanisms are extremely difficult to study experimentally and, in many cases, large-scale model networks are of great help. However, the implementation of these models relies on experimental findings that are often sparse and limited. Their predicting power ultimately depends on how closely a model’s connectivity represents the real system. Here we argue that the data-driven probabilistic rules, widely used to build neuronal network models, may not be appropriate to represent the dynamics of the corresponding biological system. To solve this problem, we propose to use a new mathematical framework able to use sparse and limited experimental data to quantitatively reproduce the structural connectivity of biological brain networks at cellular level.

Download Full-text

Reconfiguration of human evolving large-scale epileptic brain networks prior to seizures: an evaluation with node centralities

Scientific Reports ◽

10.1038/s41598-020-78899-7 ◽

2020 ◽

Vol 10 (1) ◽

Author(s):

Rieke Fruengel ◽

Timo Bröhl ◽

Thorsten Rings ◽

Klaus Lehnertz

Keyword(s):

Large Scale ◽

Brain Networks ◽

Brain Regions ◽

Brain Network ◽

Theoretical Concept ◽

Global Network ◽

Time Resolved ◽

Functional Connections ◽

Node Centrality ◽

Network Mechanism

AbstractPrevious research has indicated that temporal changes of centrality of specific nodes in human evolving large-scale epileptic brain networks carry information predictive of impending seizures. Centrality is a fundamental network-theoretical concept that allows one to assess the role a node plays in a network. This concept allows for various interpretations, which is reflected in a number of centrality indices. Here we aim to achieve a more general understanding of local and global network reconfigurations during the pre-seizure period as indicated by changes of different node centrality indices. To this end, we investigate—in a time-resolved manner—evolving large-scale epileptic brain networks that we derived from multi-day, multi-electrode intracranial electroencephalograpic recordings from a large but inhomogeneous group of subjects with pharmacoresistant epilepsies with different anatomical origins. We estimate multiple centrality indices to assess the various roles the nodes play while the networks transit from the seizure-free to the pre-seizure period. Our findings allow us to formulate several major scenarios for the reconfiguration of an evolving epileptic brain network prior to seizures, which indicate that there is likely not a single network mechanism underlying seizure generation. Rather, local and global aspects of the pre-seizure network reconfiguration affect virtually all network constituents, from the various brain regions to the functional connections between them.

Download Full-text

Static and dynamic functional connectivity supports the configuration of brain networks associated with creative cognition

Scientific Reports ◽

10.1038/s41598-020-80293-2 ◽

2021 ◽

Vol 11 (1) ◽

Author(s):

Abhishek Uday Patil ◽

Sejal Ghate ◽

Deepa Madathil ◽

Ovid J. L. Tzeng ◽

Hsu-Wen Huang ◽

...

Keyword(s):

Cognitive Processes ◽

Large Scale ◽

Brain Networks ◽

Blood Oxygen Level Dependent ◽

Detection Algorithm ◽

Brain Regions ◽

Creative Cognition ◽

Network Flexibility ◽

Community Detection Algorithm ◽

Window Approach

AbstractCreative cognition is recognized to involve the integration of multiple spontaneous cognitive processes and is manifested as complex networks within and between the distributed brain regions. We propose that the processing of creative cognition involves the static and dynamic re-configuration of brain networks associated with complex cognitive processes. We applied the sliding-window approach followed by a community detection algorithm and novel measures of network flexibility on the blood-oxygen level dependent (BOLD) signal of 8 major functional brain networks to reveal static and dynamic alterations in the network reconfiguration during creative cognition using functional magnetic resonance imaging (fMRI). Our results demonstrate the temporal connectivity of the dynamic large-scale creative networks between default mode network (DMN), salience network, and cerebellar network during creative cognition, and advance our understanding of the network neuroscience of creative cognition.

Download Full-text

CNTM-01. Evaluating Traditional and Non-Traditional Eloquent Areas in Patients with Brain Tumors: Large-scale Network Analysis Using a Machine Learning-Based Platform

Neuro-Oncology ◽

10.1093/neuonc/noab196.899 ◽

2021 ◽

Vol 23 (Supplement_6) ◽

pp. vi224-vi224

Author(s):

Alexis Morell ◽

Daniel Eichberg ◽

Ashish Shah ◽

Evan Luther ◽

Victor Lu ◽

...

Keyword(s):

Machine Learning ◽

Brain Tumors ◽

Large Scale ◽

Diffusion Tensor ◽

Brain Networks ◽

Tumor Resection ◽

Central Executive ◽

Default Mode ◽

Eloquent Areas ◽

Neurologic Deficits

Abstract BACKGROUND Developing mapping tools that allow identification of traditional or non-traditional eloquent areas is necessary to minimize the risk of postoperative neurologic deficits. The objective of our study is to evaluate the use of a novel cloud-based platform that uses machine learning to identify cerebral networks in patients with brain tumors. METHODS We retrospectively included all adult patients who underwent surgery for brain tumor resection or thermal ablation at our Institution between the 16th of February and the 15th of May of 2021. Pre and postoperative contrast-enhanced MRI with T1-weighted and high-resolution Diffusion Tensor Imaging (DTI) sequences were uploaded into the Quicktome platform. After processing the data, we categorized the integrity of seven large-scale brain networks: sensorimotor, visual, ventral attention, central executive, default mode, dorsal attention and limbic. Affected networks were correlated with pre and postoperative clinical data, including neurologic deficits. RESULTS Thirty-five (35) patients were included in the study. The average age of the sample was 63.2 years, and 51.4% (n=18) were females. The most affected network was the central executive network (40%), followed by the dorsal attention and default mode networks (31.4%), while the least affected were the visual (11%) and ventral attention networks (17%). Patients with preoperative deficits showed a significantly higher number of altered networks before the surgery (p=0.021), compared to patients without deficits. In addition, we found that patients without neurologic deficits had an average of 2.06 large-scale networks affected, with 75% of them not being related to traditional eloquent areas as the sensorimotor, language or visual circuits. CONCLUSIONS The Quicktome platform is a practical tool that allows automatic visualization of large-scale brain networks in patients with brain tumors. Although further studies are needed, it may assist in the surgical management of traditional and non-traditional eloquent areas.

Download Full-text

A Computational Network Control Theory Analysis of Depression Symptoms

Personality Neuroscience ◽

10.1017/pen.2018.15 ◽

2018 ◽

Vol 1 ◽

Cited By ~ 2

Author(s):

Yoed N. Kenett ◽

Roger E. Beaty ◽

John D. Medaglia

Keyword(s):

Control Theory ◽

Diffusion Tensor ◽

Brain Regions ◽

Brain Network ◽

Network Control ◽

Depression Symptoms ◽

Theory Approach ◽

Imaging Data ◽

The Brain ◽

Subclinical Depression

AbstractRumination and impaired inhibition are considered core characteristics of depression. However, the neurocognitive mechanisms that contribute to these atypical cognitive processes remain unclear. To address this question, we apply a computational network control theory approach to structural brain imaging data acquired via diffusion tensor imaging in a large sample of participants, to examine how network control theory relates to individual differences in subclinical depression. Recent application of this theory at the neural level is built on a model of brain dynamics, which mathematically models patterns of inter-region activity propagated along the structure of an underlying network. The strength of this approach is its ability to characterize the potential role of each brain region in regulating whole-brain network function based on its anatomical fingerprint and a simplified model of node dynamics. We find that subclinical depression is negatively related to higher integration abilities in the right anterior insula, replicating and extending previous studies implicating atypical switching between the default mode and Executive Control Networks in depression. We also find that subclinical depression is related to the ability to “drive” the brain system into easy to reach neural states in several brain regions, including the bilateral lingual gyrus and lateral occipital gyrus. These findings highlight brain regions less known in their role in depression, and clarify their roles in driving the brain into different neural states related to depression symptoms.

Download Full-text

Group-level progressive alterations in brain connectivity patterns revealed by diffusion-tensor brain networks across severity stages in Alzheimer’s disease

10.1101/105270 ◽

2017 ◽

Author(s):

J. Rasero ◽

C. Alonso-Montes ◽

I. Diez ◽

L. Olabarrieta-Landa ◽

L. Remaki ◽

...

Keyword(s):

Alzheimer’S Disease ◽

Alzheimer's Disease ◽

Cognitive Impairment ◽

Mild Cognitive Impairment ◽

Disease Severity ◽

Diffusion Tensor ◽

Brain Networks ◽

Stage Ii ◽

Group Level ◽

The Brain

AbstractAlzheimer’s disease (AD) is a chronically progressive neurodegenerative disease highly correlated to aging. Whether AD originates by targeting a localized brain area and propagates to the rest of the brain across disease-severity progression is a question with an unknown answer. Here, we aim to provide an answer to this question at the group-level by looking at differences in diffusion-tensor brain networks. In particular, making use of data from Alzheimer's Disease Neuroimaging Initiative (ADNI), four different groups were defined (all of them matched by age, sex and education level): G1 (N1=36, healthy control subjects, Control), G2 (N2=36, early mild cognitive impairment, EMCI), G3 (N3=36, late mild cognitive impairment, LMCI) and G4 (N4=36, AD). Diffusion-tensor brain networks were compared across three disease stages: stage I 3(Control vs EMCI), stage II (Control vs LMCI) and stage III (Control vs AD). The group comparison was performed using the multivariate distance matrix regression analysis, a technique that was born in genomics and was recently proposed to handle brain functional networks, but here applied to diffusion-tensor data. The results were three-fold: First, no significant differences were found in stage I. Second, significant differences were found in stage II in the connectivity pattern of a subnetwork strongly associated to memory function (including part of the hippocampus, amygdala, entorhinal cortex, fusiform gyrus, inferior and middle temporal gyrus, parahippocampal gyrus and temporal pole). Third, a widespread disconnection across the entire AD brain was found in stage III, affecting more strongly the same memory subnetwork appearing in stage II, plus the other new subnetworks,including the default mode network, medial visual network, frontoparietal regions and striatum. Our results are consistent with a scenario where progressive alterations of connectivity arise as the disease severity increases and provide the brain areas possibly involved in such a degenerative process. Further studies applying the same strategy to longitudinal data are needed to fully confirm this scenario.

Download Full-text

Graph Theory-Based Brain Network Connectivity Analysis and Classification of Alzheimer’s Disease

International Journal of Image and Graphics ◽

10.1142/s021946782240006x ◽

2021 ◽

Author(s):

A. Thushara ◽

C. Ushadevi Amma ◽

Ansamma John

Keyword(s):

Alzheimer’S Disease ◽

Alzheimer's Disease ◽

Graph Theory ◽

Neurodegenerative Disorder ◽

Brain Networks ◽

Network Connectivity ◽

Brain Regions ◽

Brain Network ◽

Fiber Tracts ◽

The Brain

Alzheimer’s Disease (AD) is basically a progressive neurodegenerative disorder associated with abnormal brain networks that affect millions of elderly people and degrades their quality of life. The abnormalities in brain networks are due to the disruption of White Matter (WM) fiber tracts that connect the brain regions. Diffusion-Weighted Imaging (DWI) captures the brain’s WM integrity. Here, the correlation betwixt the WM degeneration and also AD is investigated by utilizing graph theory as well as Machine Learning (ML) algorithms. By using the DW image obtained from Alzheimer’s Disease Neuroimaging Initiative (ADNI) database, the brain graph of each subject is constructed. The features extracted from the brain graph form the basis to differentiate between Mild Cognitive Impairment (MCI), Control Normal (CN) and AD subjects. Performance evaluation is done using binary and multiclass classification algorithms and obtained an accuracy that outperforms the current top-notch DWI-based studies.

Download Full-text

Altered topology of large-scale structural brain networks in chronic stroke

Brain Communications ◽

10.1093/braincomms/fcz020 ◽

2019 ◽

Vol 1 (1) ◽

Cited By ~ 1

Author(s):

Bastian Cheng ◽

Eckhard Schlemm ◽

Robert Schulz ◽

Marlene Boenstrup ◽

Arnaud Messé ◽

...

Keyword(s):

Information Transfer ◽

Large Scale ◽

Neuronal Degeneration ◽

Brain Networks ◽

Brain Regions ◽

Local Network ◽

Structural Connectome ◽

The Impact ◽

Structural Brain ◽

Topological Changes

Abstract Beyond disruption of neuronal pathways, focal stroke lesions induce structural disintegration of distant, yet connected brain regions via retrograde neuronal degeneration. Stroke lesions alter functional brain connectivity and topology in large-scale brain networks. These changes are associated with the degree of clinical impairment and recovery. In contrast, changes of large scale, structural brain networks after stroke are less well reported. We therefore aimed to analyse the impact of focal lesions on the structural connectome after stroke based on data from diffusion-weighted imaging and probabilistic fibre tracking. In total, 17 patients (mean age 64.5 ± 8.4 years) with upper limb motor deficits in the chronic stage after stroke and 21 healthy participants (mean age 64.9 ± 10.3 years) were included. Clinical deficits were evaluated by grip strength and the upper extremity Fugl-Meyer assessment. We calculated global and local graph theoretical measures to characterize topological changes in the structural connectome. Results from our analysis demonstrated significant alterations of network topology in both ipsi- and contralesional, primarily unaffected, hemispheres after stroke. Global efficiency was significantly lower in stroke connectomes as an indicator of overall reduced capacity for information transfer between distant brain areas. Furthermore, topology of structural connectomes was shifted toward a higher degree of segregation as indicated by significantly higher values of global clustering and modularity. On a level of local network parameters, these effects were most pronounced in a subnetwork of cortico-subcortical brain regions involved in motor control. Structural changes were not significantly associated with clinical measures. We propose that the observed network changes in our patients are best explained by the disruption of inter- and intrahemispheric, long white matter fibre tracts connecting distant brain regions. Our results add novel insights on topological changes of structural large-scale brain networks in the ipsi- and contralesional hemisphere after stroke.

Download Full-text