The growing availability of well-organized Electronic Health Records (EHR) data has enabled the development of various machine learning models towards disease risk prediction. However, existing risk prediction methods overlook the heterogeneity of complex diseases, failing to model the potential disease subtypes regarding their corresponding patient visits and clinical concept subgroups. In this work, we introduce TACCO, a novel framework that jointly discovers clusters of clinical concepts and patient visits based on a hypergraph modeling of EHR data. Specifically, we develop a novel self-supervised co-clustering framework that can be guided by the risk prediction task of specific diseases. Furthermore, we enhance the hypergraph model of EHR data with textual embeddings and enforce the alignment between the clusters of clinical concepts and patient visits through a contrastive objective. Comprehensive experiments conducted on the public MIMIC-III dataset and Emory internal CRADLE dataset over the downstream clinical tasks of phenotype classification and cardiovascular risk prediction demonstrate an average 31.25% performance improvement compared to traditional ML baselines and a 5.26% improvement on top of the vanilla hypergraph model without our co-clustering mechanism. In-depth model analysis, clustering results analysis, and clinical case studies further validate the improved utilities and insightful interpretations delivered by TACCO. Code is available at https://github.com/PericlesHat/TACCO.

Biological networks are commonly used in biomedical and healthcare domains to effectively model the structure of complex biological systems with interactions linking biological entities. However, due to their characteristics of high dimensionality and low sample size, directly applying deep learning models on biological networks usually faces severe overfitting. In this work, we propose R-MIXUP, a Mixup-based data augmentation technique that suits the symmetric positive definite (SPD) property of adjacency matrices from biological networks with optimized training efficiency. The interpolation process in R-MIXUP leverages the log-Euclidean distance metrics from the Riemannian manifold, effectively addressing the swelling effect and arbitrarily incorrect label issues of vanilla Mixup. We demonstrate the effectiveness of R-MIXUP with five real-world biological network datasets on both regression and classification tasks. Besides, we derive a commonly ignored necessary condition for identifying the SPD matrices of biological networks and empirically study its influence on the model performance. The code implementation can be found in Appendix E.

Mapping the connectome of the human brain using structural or functional connectivity has become one of the most pervasive paradigms for neuroimaging analysis. Recently, Graph Neural Networks (GNNs) motivated from geometric deep learning have attracted broad interest due to their established power for modeling complex networked data. Despite their superior performance in many fields, there has not yet been a systematic study of how to design effective GNNs for brain network analysis. To bridge this gap, we present BrainGB, a benchmark for brain network analysis with GNNs. BrainGB standardizes the process by (1) summarizing brain network construction pipelines for both functional and structural neuroimaging modalities and (2) modularizing the implementation of GNN designs. We conduct extensive experiments on datasets across cohorts and modalities and recommend a set of general recipes for effective GNN designs on brain networks. To support open and reproducible research on GNN-based brain network analysis, we host the BrainGBwebsite at https://braingb.us with models, tutorials, examples, as well as an out-of-box Python package. We hope that this work will provide useful empirical evidence and offer insights for future research in this novel and promising direction.

Brain networks, graphical models such as those constructed from MRI, have been widely used in pathological prediction and analysis of brain functions. Within the complex brain system, differences in neuronal connection strengths parcellate the brain into various functional modules (network communities), which are critical for brain analysis. However, identifying such communities within the brain has been a nontrivial issue due to the complexity of neuronal interactions. In this work, we propose a novel interpretable transformer-based model for joint hierarchical cluster identification and brain network classification with three main contributions. First, we offer an end-to-end transformer-based approach to learning clustering assignments. Through pairwise attention, a clustering layer, BCluster, and a transformer encoder collaboratively learn a globally shared clustering assignment that is continuously tuned to downstream tasks. BCluster enhances the model’s performance and reduces run time complexity while also providing clinical insights. Second, we propose a hierarchical structure for the clustering model, enabling the model to learn more abstract, higher-level cluster representations by combining lower-level modules. Each clustering layer is attached to a distinct readout module, which allows the model to utilize the cluster embeddings of every layer effectively. Last but not least, we redesign the attention mechanism of the transformer with stochastic noise, which enhances its cluster learning capability. We compare our model’s performance with SOTA models and perform clustering analysis with the ground truth community labels. Extensive experimental results show that with the help of hierarchical clustering, the model achieves increased accuracy and reduced runtime complexity while providing plausible insight into the functional organization of brain regions.

Human brains, controlling behaviors and cognition, are at the center of complex neurobiological systems. Recent studies in neuroscience and neuroimaging analysis have reached a consensus that interactions among brain regions of interest (ROIs) are driving factors for neural development and disorders. Graph neural networks (GNNs) as a powerful tool for analyzing graph-structured data are naturally applied to the analysis of brain networks. However, training of deep learning models including GNNs often requires a significant amount of labeled data. Due to the complicated data acquisition process and restrictions on data sharing, brain network datasets are still small compared to other types of graphs (e.g., social networks, molecules, proteins). Moreover, real clinical tasks (e.g., mental disorder analysis) are often conducted on local datasets with even smaller scales and larger noises. To this end, we propose to leverage pre-training to capture the intrinsic brain network structures regardless of specific clinical outcomes, for obtaining GNN models that are easily adaptable to downstream tasks. Specifically, (1) we design brain-network-oriented unsupervised pre-training techniques to utilize large-scale brain imaging studies without highly relevant task labels; (2) we develop a data-driven parcellation atlas mapping pipeline to facilitate effective knowledge transfer across studies with different ROI systems. The proposed framework is validated with various GNN models, with extensive empirical results demonstrating consistent improvement in performance as well as robustness.

Functional magnetic resonance imaging (fMRI) has become one of the most common imaging modalities for brain function analysis. Recently, graph neural networks (GNN) have been adopted for fMRI analysis with superior performance. Unfortunately, traditional functional brain networks are mainly constructed based on similarities among region of interests (ROI), which are noisy and agnostic to the downstream prediction tasks and can lead to inferior results for GNN-based models. To better adapt GNNs for fMRI analysis, we propose TBDS, an end-to-end framework based on Task-aware Brain connectivity DAG (short for Directed Acyclic Graph) Structure generation for fMRI analysis. The key component of TBDS is the brain network generator which adopts a DAG learning approach to transform the raw time-series into task-aware brain connectivities. Besides, we design an additional contrastive regularization to inject task-specific knowledge during the brain network generation process. Comprehensive experiments on two fMRI datasets, namely Adolescent Brain Cognitive Development (ABCD) and Philadelphia Neuroimaging Cohort (PNC) datasets demonstrate the efficacy of TBDS. In addition, the generated brain networks also highlight the prediction-related brain regions and thus provide unique interpretations of the prediction results. Our implementation will be published upon acceptance.

Graph neural networks (GNNs) have been widely used in various graph-related problems such as node classification and graph classification, where the superior performance is mainly established when natural node features are available. However, it is not well understood how GNNs work without natural node features, especially regarding the various ways to construct artificial ones. In this paper, we point out the two types of artificial node features, i.e., positional and structural node features, and provide insights on why each of them is more appropriate for certain tasks, i.e., positional node classification, structural node classification, and graph classification. Extensive experimental results on 10 benchmark datasets validate our insights, thus leading to a practical guideline on the choices between different artificial node features for GNNs on non-attributed graphs. The code is available at https://github.com/zjzijielu/gnn-positional-structural-node-features.

Brain networks characterize complex connectivities among brain regions as graph structures, which provide a powerful means to study brain connectomes. In recent years, graph neural networks have emerged as a prevalent paradigm of learning with structured data. However, most brain network datasets are limited in sample sizes due to the relatively high cost of data collection, which hinders the deep learning models from sufficient training. Inspired by meta-learning that learns new concepts fast with limited training examples, this paper studies data-efficient training strategies for analyzing brain connectomes in a cross-dataset setting. Specifically, we propose to meta-train the model on datasets of large sample sizes and transfer the knowledge to small datasets. In addition, we also explore brain-network-oriented designs, including atlas transformation and adaptive task reweighing. Compared to other pre-training strategies, our meta-learning-based approach achieves higher and stabler performance, which demonstrates the effectiveness of our proposed solutions. The framework is also able to derive new insights regarding the similarities among datasets and diseases in a data-driven fashion.

Recent years have seen a growing interest in Graph Contrastive Learning (GCL), which trains Graph Neural Network (GNN) model to discriminate similar and dissimilar pairs of nodes without human annotations. Most prior GCL work focuses on homogeneous graphs and little attention has been paid to Heterogeneous Graphs (HGs) that involve different types of nodes and edges. Moreover, earlier studies reveal that the explicit use of structure information of underlying graphs is useful for learning representations. Conventional GCL methods merely measure the likelihood of contrastive pairs according to node representations, which may not align with the true semantic similarities. How to leverage such structure information for GCL is not yet well-understood. To address the aforementioned challenges, this paper presents a novel method dubbed STructure-EnhaNced heterogeneous graph ContrastIve Learning, STENCIL for brevity. At first, we generate multiple semantic views for HGs based on metapaths. Unlike most methods that maximize the consistency among these views, we propose a novel multiview contrastive aggregation objective to adaptively distill information from each view. In addition, we advocate the explicit use of structure embedding, which enriches the model with local structural patterns of the underlying HGs, so as to better mine true and hard negatives for GCL. Empirical studies on three real-world datasets show that our proposed method consistently outperforms existing state-of-the-art methods and even surpasses several supervised counterparts.