Spatio-Temporal Graph Convolution for Resting-State fMRI Analysis

Gadgil, Soham; Zhao, Qingyu; Pfefferbaum, Adolf; Sullivan, Edith V.; Adeli, Ehsan; Pohl, Kilian M.

doi:10.1007/978-3-030-59728-3_52

Cited by 100 publications

(103 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A wide range of methods from other research domains have been adapted to study brain connectivity, including the Spatio-Temporal Graph Convolution (ST-CGN) [38] which took advantage of a graph convolutional network to model the nonstationary properties of functional connectivity by learning the importance of edges between brain networks and showed 80 percent accuracy in predicting gender and more than 75 percent on predicting age using the encoded temporal dynamics. However, this model, like most others, did not include spatial variability as a consequence of using a predefined atlas for extracting ROIs.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

A novel 5D brain parcellation approach based on spatio-temporal encoding of resting fMRI data from deep residual learning

Kazemivash

2021

Preprint

View full text Add to dashboard Cite

ObjectiveBrain parcellation is an essential aspect of computational neuroimaging research and deals with segmenting the brain into (possibly overlapping) sub-regions employed to study brain anatomy or function. In the context of functional parcellation, brain organization which is often measured via temporal metrics such as coherence, is highly dynamic. This dynamic aspect is ignored in most research, which typically applies anatomically based, fixed regions for each individual, and can produce misleading results.MethodsIn this work, we propose a novel spatio-temporal-network (5D) brain parcellation scheme utilizing a deep residual network to predict the probability of each voxel belonging to a brain network at each point in time.ResultsWe trained 53 4D brain networks and evaluate the ability of these networks to capture spatial and temporal dynamics as well as to show sensitivity to individual or group-level variation (in our case with age).ConclusionThe proposed system generates informative spatio-temporal networks that vary not only across individuals but also over time and space.SignificanceThe dynamic 5D nature of the developed approach provides a powerful framework that expands on existing work and has potential to identify novel and typically ignored findings when studying the healthy and disordered brain.

show abstract

Section: Discussionmentioning

confidence: 99%

“…More recently there has been recognition the need for also modeling changes in the spatial nodes/networks over time as well [17,35] including changes in shape, size or translation of active regions. If we consider both spatial and temporal features for a given subject simultaneously then spatio-temporal dynamics is taken into account [38].…”

Section: Introductionmentioning

confidence: 99%

A novel 5D brain parcellation approach based on spatio-temporal encoding of resting fMRI data from deep residual learning

Kazemivash

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…In contrast, spatial GCNs imitate the Euclidean convolution on grid data to aggregate spatial features between neighboring nodes. Although spectral GCNs have achieved great success on both structural and functional MRI applications (Gopinath, Desrosiers, & Lombaert, 2019 ; Hong et al, 2019 ; Ktena et al, 2018 ; Parisot et al, 2017 ), spatial models are preferred over the spectral ones because of their efficiency, generalization, and flexibility (Monti et al, 2017 ; Wu et al, 2019 ; Zhang, Cui, & Zhu, 2018 ), and they have gained increasing interest in the community (Azevedo, Passamonti, Liò, & Toschi, 2020 ; Gadgil et al, 2020 ).…”

Section: Introductionmentioning

confidence: 99%

Graph convolutional network for fMRI analysis based on connectivity neighborhood

Wang

2021

Network Neuroscience

View full text Add to dashboard Cite

There have been successful applications of deep learning to functional magnetic resonance imaging (fMRI), where fMRI data were mostly considered as structured grids, and spatial features from Euclidean neighbors were usually extracted by the convolutional neural networks (CNN) in the computer vision field. Recently, CNN has been extended to graph data and demonstrated superior performance. Here, we define graphs based on functional connectivity and present a connectivity-based graph convolution network (cGCN) architecture for fMRI analysis. Such an approach allows us to extract spatial features from connectomic neighborhoods rather than Euclidean ones, consistent with the functional organization of the brain. To evaluate the performance of cGCN, we applied it to two scenarios with resting-state fMRI data. One is individual identification of healthy participants and the other is classification of Autistic patients from normal controls. Our results indicate that cGCN can effectively capture functional connectivity features in fMRI analysis for relevant applications.

show abstract

“…In response to these challenges, numerous studies have sought to model neuroimaging data, most typically focusing on widely available benchmark datasets using resting state connectivity Suk et al, 2016;Mao et al, 2019;Tahmassebi et al, 2018), structural (Henschel et al, 2020;Tian et al, 2020;Gunawardena et al, 2017;Zhang, 2018) and fMRI (Gadgil et al, 2020;Riaz et al, 2018;Sarraf and Tofighi, 2016) modalities. However, this large and growing literature, which most often focuses on prediction and classification, has largely ignored task-based fMRI, data that result from designed experiments aimed at testing particular scientific hypotheses.…”

Section: Introductionmentioning

confidence: 99%

Deep Generative Analysis for Task-Based Functional MRI Experiments

D¹,

Goffinet²,

Wright

2021

Preprint

View full text Add to dashboard Cite

While functional magnetic resonance imaging (fMRI) remains one of the most widespread and important methods in basic and clinical neuroscience, the data it produces---time series of brain volumes---continue to pose daunting analysis challenges. The current standard ("mass univariate") approach involves constructing a matrix of task regressors, fitting a separate general linear model at each volume pixel ("voxel"), computing test statistics for each model, and correcting for false positives post hoc using bootstrap or other resampling methods. Despite its simplicity, this approach has enjoyed great success over the last two decades due to: 1) its ability to produce effect maps highlighting brain regions whose activity significantly correlates with a given variable of interest; and 2) its modeling of experimental effects as separable and thus easily interpretable. However, this approach suffers from several well-known drawbacks, namely: inaccurate assumptions of linearity and noise Gaussianity; a limited ability to capture individual effects and variability; and difficulties in performing proper statistical testing secondary to independently fitting voxels. In this work, we adopt a different approach, modeling entire volumes directly in a manner that increases model flexibility while preserving interpretability. Specifically, we use a generalized additive model (GAM) in which the effects of each regressor remain separable, the product of a spatial map produced by a variational autoencoder and a (potentially nonlinear) gain modeled by a covariate-specific Gaussian Process. The result is a model that yields group-level effect maps comparable or superior to the ones obtained with standard fMRI analysis software while also producing single-subject effect maps capturing individual differences. This suggests that generative models with a decomposable structure might offer a more flexible alternative for the analysis of task-based fMRI data.

show abstract

Spatio-Temporal Graph Convolution for Resting-State fMRI Analysis

Cited by 100 publications

References 28 publications

A novel 5D brain parcellation approach based on spatio-temporal encoding of resting fMRI data from deep residual learning

A novel 5D brain parcellation approach based on spatio-temporal encoding of resting fMRI data from deep residual learning

Graph convolutional network for fMRI analysis based on connectivity neighborhood

Deep Generative Analysis for Task-Based Functional MRI Experiments

Contact Info

Product

Resources

About