BrainGB: A Benchmark for Brain Network Analysis With Graph Neural Networks

Cui, Hejie; Dai, Wei; Zhu, Yanqiao; Kan, Xuan; Gu, Antonio Aodong Chen; Lukemire, Joshua; Zhan, Liang; He, Lifang; Guo, Ying; Yang, Carl

doi:10.1109/tmi.2022.3218745

Cited by 63 publications

(27 citation statements)

References 83 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One step further, when GNNs are used for brain network analysis, a graph-level embedding needs to be generated through a readout function based on the learned node embeddings [37,43,13]. As is shown in Figure 1(a), a property of brain networks is that brain regions (nodes) belonging to the same functional modules often share similar behaviors regarding activations and deactivations in response to various stimulations [7].…”

Section: Orthonormal Bases Non-orthonormal Basesmentioning

confidence: 99%

“…In addition, FBNetGen [35] considers the learnable generation of brain networks and explores the explainability of the generated brain networks towards downstream tasks. Another benchmark paper [13] systematically studies the effectiveness of various GNN designs over brain network data. Different from other work focusing on static brain networks, STAGIN [39] utilizes GNNs with spatio-temporal attention to model dynamic brain networks extracted from fMRI data.…”

Section: Gnns For Brain Network Analysismentioning

confidence: 99%

“…However, brain networks have several unique traits that make directly applying existing graph Transformer models impractical. First, one of the simplest and most frequently used methods to construct a brain network in the neuroimaging community is via pairwise correlations between BOLD time courses from two ROIs [43,35,13,63,69]. This impedes the designs like centrality, spatial, and edge encoding because each node in the brain network has the same degree and connects to every other node by a single hop.…”

Section: Introductionmentioning

confidence: 99%

“…Second, in previous graph transformer models, eigenvalues and eigenvectors are commonly used as positional embeddings because they can provide identity and positional information for each node [15,26]. Nevertheless, in brain networks, the connection profile, which is defined as each node's corresponding row in the brain network adjacency matrix, is recognized as the most effective node feature [13]. This node feature naturally encodes both structural and positional information, making the aforementioned positional embedding design based on eigenvalues and eigenvectors redundant.…”

Section: Introductionmentioning

confidence: 99%

“…In this work, we propose to develop BRAIN NETWORK TRANSFORMER (BRAINNETTF), which leverages the unique properties of brain network data to fully unleash the power of Transformer-based models for brain network analysis. Specifically, motivated by previous findings on effective GNN designs for brain networks [13], we propose to use the effective initial node features of connection profiles. Empirical analysis shows that connection profiles naturally provide positional features for Transformer-based models and avoid the costly computations of eigenvalues or eigenvectors.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Brain Network Transformer

Kan¹,

Dai²,

Cui³

et al. 2022

Preprint

View full text Add to dashboard Cite

Human brains are commonly modeled as networks of Regions of Interest (ROIs) and their connections for the understanding of brain functions and mental disorders. Recently, Transformer-based models have been studied over different types of data, including graphs, shown to bring performance gains widely. In this work, we study Transformer-based models for brain network analysis. Driven by the unique properties of data, we model brain networks as graphs with nodes of fixed size and order, which allows us to (1) use connection profiles as node features to provide natural and low-cost positional information and (2) learn pairwise connection strengths among ROIs with efficient attention weights across individuals that are predictive towards downstream analysis tasks. Moreover, we propose an ORTHONORMAL CLUSTERING READOUT operation based on selfsupervised soft clustering and orthonormal projection. This design accounts for the underlying functional modules that determine similar behaviors among groups of ROIs, leading to distinguishable cluster-aware node embeddings and informative graph embeddings. Finally, we re-standardize the evaluation pipeline on the only one publicly available large-scale brain network dataset of ABIDE, to enable meaningful comparison of different models. Experiment results show clear improvements of our proposed BRAIN NETWORK TRANSFORMER on both the public ABIDE and our restricted ABCD datasets. The implementation is available at https://github.com/Wayfear/BrainNetworkTransformer. Nowadays Transformer-based models have led a tremendous success in various downstream tasks across fields including natural language processing [56, 17] and computer vision [20, 10, 55]. Recent efforts have also emerged to apply Transformer-based designs to graph representation learning. GAT [57] firstly adapts the attention mechanism to graph neural networks (GNNs) but only considers the local structures of neighboring nodes. Graph Transformer [21] injects edge information into the attention mechanism and leverages the eigenvectors of each node as positional embeddings. SAN [40] further enhances the positional embeddings by considering both eigenvalues and eigenvectors and improves the attention mechanism by extending the attention from local to global structures.36th Conference on Neural Information Processing Systems (NeurIPS 2022).

show abstract