Hongmin Cai

AAAI Conference 2026 Conference Paper

Graph Contrastive Learning with Balanced Hard Negatives and Fine-grained Semantic-aware Positives

• Hongshan Pu
• Haoxu Zhang
• Ye Liu
• Hongmin Cai

Graph contrastive learning (GCL) aims to learn representations by bringing semantically similar graphs closer and pushing dissimilar ones farther apart without label supervision. Hard negatives, which refer to graphs that have different labels but similar embeddings to the target graph, play a key role in improving representation discrimination. However, current methods that generate both high-quality positives and hard negatives face two challenges: (1) Hard negative sample generation often suffers from class imbalance, resulting in unequal attention across classes and reduced discriminative power in the learned representations. (2) The typical binary positive sample generation approach, which divides the graph into important and unimportant semantic regions, overlooks regions that negatively impact semantics and mislead model predictions. To address these issues, we introduce a novel method named BalanceGCL, which enhance graph contrastive learning with balanced hard negatives and fine-grained semantic-aware positives. BalanceGCL comprises two modules: Balanced Hard Negative graphs generation (BHN) and Fine-grained Semantic-aware Positive graphs generation (FSP). Inspired by the counterfactual mechanism, BHN generates balanced hard negatives that remain structurally similar to the original graph while inducing a controlled semantic shift. To ensure class balance, BHN iteratively constructs one hard negative sample for each class, ensuring an even distribution of negative samples across all alternative categories. FSP leverages the semantic differences between original graphs and balanced hard negatives to identify positively contributing, negatively contributing, and unimportant regions. By enhancing the influence of positive contributors, suppressing negative ones, and perturbing unimportant areas, it generates more reliable and semantically complete positive samples. The proposed method outperforms state-of-the-art GCL techniques across 14 datasets in graph classification and transfer learning tasks, demonstrating its effectiveness in tackling class imbalance and identifying fine-grained semantic-aware regions.

JBHI Journal 2026 Journal Article

GraphSTAR: Proximal Operator-Based Graph Neural Network Enhanced by Dynamic Graph Aggregation for Spatial Transcriptomics

• Junyu Li
• Jingquan Yan
• Yi Liao
• Wenxiong Liao
• Ye Liu
• Hongmin Cai

Spatial transcriptomics technologies carry out advanced sequencing analysis of molecular profiles with a spatial context, providing multi-source information essential for elucidating biological regulatory mechanisms. Nonetheless, it poses challenges in the integration of raw spatial coordinates with high-dimensional gene expression profiles in their native feature space. While spatial-aware methods effectively aggregate molecular information from local spatial neighborhoods, they fail to explore the long-range relationships associated with gene expression data. To address this issue, this paper introduces a novel approach termed GraphSTAR that encodes both spatial and gene expression data into undirected graphs, characterizing the local spatial proximity and global transcriptional similarity, respectively. Through a graph aggregation process, GraphSTAR integrates these diverse data sources within a joint graph structure, effectively modeling both local neighborhood relationships and long-range functional associations. Subsequently, a reassembled graph neural network is established by incorporating the graph aggregation into the feed-forward propagation using proximal operators, progressively refining spatial-informed latent representation to decipher spatial expression patterns of genes. Extensive experiments on benchmark datasets demonstrate that GraphSTAR outperforms state-of-the-art methods in both spatial domain identification and cell-type annotation tasks.

AAAI Conference 2026 Conference Paper

MIGDiff: Multi-attributes Imputations for Attribute-missing Graphs via Graph Denoising Diffusion Model

• Ye Liu
• Yang Chen
• Hongmin Cai

The missing of graph attributes poses a significant challenge in graph representation learning. Some existing graph attribute completion methods adopt the shared-space hypothesis or employ end-to-end frameworks to perform single-attribute imputation. However, these models can only generate one single attribute with a few specific patterns that either adhere to prior knowledge or are optimal for downstream tasks, making it difficult to capture the full range of variations in the target attribute distribution. This limitation negatively impacts the model's generalizability and efficiency. Therefore, to address this issue, we proposed a new method based on a graph denoising diffusion model, called Multi-attribute Imputation Graph Denoising Diffusion Model (MIGDiff), which can generate multiple high-quality attributes. Specifically, it employs a Dual-source Auto-encoder on existing attributes and graph topology to extract reliable knowledge, which serves as a condition for training the diffusion module. Within diffusion, noise is added to the structural embeddings of nodes without attributes in the forward process. In the reverse process, a Structure-aware Denoising Network is devised to integrate feature and structural information via an attention mechanism and to perform neighbor-guided refinement based on graph connectivity, thereby enhancing denoising and accurately recovering missing attributes while effectively maintaining structural consistency and distributional fidelity. During generation, multiple initial values are sampled to produce diverse attribute imputations, avoiding focusing on a few easy-to-learn patterns. Extensive experiments conducted on four public datasets highlight the state-of-the-art performance of MIGDiff in both attribute imputation and node classification tasks.

JBHI Journal 2025 Journal Article

Harmonic Wavelet Neural Network for Discovering Neuropathological Propagation Patterns in Alzheimer's Disease

• Hongmin Cai
• Ranran Deng
• Defu Yang
• Fa Zhang
• Guorong Wu
• Jiazhou Chen

Emerging researchindicates that the degenerative biomarkers associated with Alzheimer's disease (AD) exhibit a non-random distribution within the cerebral cortex, instead following the structural brain network. The alterations in brain networks occur much earlier than the onset of clinical symptoms, thereby affecting the progression of brain disease. In this context, the utilization of computational methods to ascertain the propagation patterns of neuropathological events would contribute to the comprehension of the pathophysiological mechanism involved in the evolution of AD. Despite the encouraging findings achieved by existing graph-based deep learning approaches in analyzing irregular graph data, their applications in identifying the spreading pathway of neuropathology are limited due to two disadvantages. They include (1) lack of a common brain network as an unbiased reference basis for group comparison, and (2) lack of an appropriate mechanism for the identification of propagation patterns. To this end, we propose a proof-of-concept harmonic wavelet neural network (HWNN) to predict the early stage of AD and localize disease-related significant wavelets, which can be used to characterize the spreading pathways of neuropathological events across the brain network. The extensive experiments constructed on both synthetic and real datasets demonstrate that our proposed method achieves superior performance in classification accuracy and statistical power of identifying propagation patterns, compared with other representative approaches.

NeurIPS Conference 2025 Conference Paper

Plug-and-play Feature Causality Decomposition for Multimodal Representation Learning

• Ye Liu
• Zihan Ji
• Hongmin Cai

Multimodal representation learning is critical for a wide range of applications, such as multimodal sentiment analysis. Current multimodal representation learning methods mainly focus on the multimodal alignment or fusion strategies, such that the complementary and consistent information among heterogeneous modalities can be fully explored. However, they mistakenly treat the uncertainty noise within each modality as the complementary information, failing to simultaneously leverage both consistent and complementary information while eliminating the aleatoric uncertainty within each modality. To address this issue, we propose a plug-and-play feature causality decomposition method for multimodal representation learning from causality perspective, which can be integrated into existing models with no affects on the original model structures. Specifically, to deal with the heterogeneity and consistency, according to whether it can be aligned with other modalities, the unimodal feature is first disentangled into two parts: modality-invariant (the synergistic information shared by all heterogeneous modalities) and modality-specific part. To deal with complementarity and uncertainty, the modality-specific part is further decomposed into unique and redundant features, where the redundant feature is removed and the unique feature is reserved based on the backdoor-adjustment. The effectiveness of noise removal is supported by causality theory. Finally, the task-related information, including both synergistic and unique components, is further fed to the original fusion module to obtain the final multimodal representations. Extensive experiments show the effectiveness of our proposed strategies.

AIIM Journal 2024 Journal Article

Corrigendum to “DeepGA for automatically estimating fetal gestational age through ultrasound imaging” [Artif. Intell. Med. 135 (2023) 102453]

• Tingting Dan
• Xijie Chen
• Miao He
• Hongmei Guo
• Xiaoqin He
• Jiazhou Chen
• Jianbo Xian
• Yu Hu

JBHI Journal 2024 Journal Article

Improving Cancer Survival Prediction via Graph Convolutional Neural Network Learning on Protein-Protein Interaction Networks

• Hongmin Cai
• Yi Liao
• Lei Zhu
• Zhikang Wang
• Jiangning Song

Cancer is one of the most challenging health problems worldwide. Accurate cancer survival prediction is vital for clinical decision making. Many deep learning methods have been proposed to understand the association between patients’ genomic features and survival time. In most cases, the gene expression matrix is fed directly to the deep learning model. However, this approach completely ignores the interactions between biomolecules, and the resulting models can only learn the expression levels of genes to predict patient survival. In essence, the interaction between biomolecules is the key to determining the direction and function of biological processes. Proteins are the building blocks and principal undertakings of life activities, and as such, their complex interaction network is potentially informative for deep learning methods. Therefore, a more reliable approach is to have the neural network learn both gene expression data and protein interaction networks. We propose a new computational approach, termed CRESCENT, which is a protein-protein interaction (PPI) prior knowledge graph-based convolutional neural network (GCN) to improve cancer survival prediction. CRESCENT relies on the gene expression networks rather than gene expression levels to predict patient survival. The performance of CRESCENT is evaluated on a large-scale pan-cancer dataset consisting of 5991 patients from 16 different types of cancers. Extensive benchmarking experiments demonstrate that our proposed method is competitive in terms of the evaluation metric of the time-dependent concordance index( ${C}^{td}$ ) when compared with several existing state-of-the-art approaches. Experiments also show that incorporating the network structure between genomic features effectively improves cancer survival prediction.

JBHI Journal 2023 Journal Article

Deep Manifold Harmonic Network With Dual Attention for Brain Disorder Classification

• Xiaoqi Sheng
• Jiazhou Chen
• Yong Liu
• Bin Hu
• Hongmin Cai

Numerous studies have shown that accurate analysis of neurological disorders contributes to the early diagnosis of brain disorders and provides a window to diagnose psychiatric disorders due to brain atrophy. The emergence of geometric deep learning approaches provides a new way to characterize geometric variations on brain networks. However, brain network data suffer from high heterogeneity and noise. Consequently, geometric deep learning methods struggle to identify discriminative and clinically meaningful representations from complex brain networks, resulting in poor diagnostic accuracy. Hence, the primary challenge in the diagnosis of brain diseases is to enhance the identification of discriminative features. To this end, this paper presents a dual-attention deep manifold harmonic discrimination (DA-DMHD) method for early diagnosis of neurodegenerative diseases. Here, a low-dimensional manifold projection is first learned to comprehensively exploit the geometric features of the brain network. Further, attention blocks with discrimination are proposed to learn a representation, which facilitates learning of group-dependent discriminant matrices to guide downstream analysis of group-specific references. Our proposed DA-DMHD model is evaluated on two independent datasets, ADNI and ADHD-200. Experimental results demonstrate that the model can tackle the hard-to-capture challenge of heterogeneous brain network topological differences and obtain excellent classifying performance in both accuracy and robustness compared with several existing state-of-the-art methods.

AIIM Journal 2023 Journal Article

DeepGA for automatically estimating fetal gestational age through ultrasound imaging

• Tingting Dan
• Xijie Chen
• Miao He
• Hongmei Guo
• Xiaoqin He
• Jiazhou Chen
• Jianbo Xian
• Yu Hu

Accurate estimation of gestational age (GA) is vital for identifying fetal abnormalities. Conventionally, GA is estimated by measuring the morphology of the cranium, abdomen, and femur manually and inputting them into the classic Hadlock formula to assess fetal growth. However, this procedure incurs considerable overhead and suffers from bias caused by the operators, yielding suboptimal estimations. To address this challenge, we develop an automatic DeepGA model to achieve fully automatic GA prediction in an end-to-end manner. Our model uses a deep segmentation model (DeepSeg) to accurately identify and segment three critical tissues, including the cranium, abdomen, and femur, in which their morphology is automatically extracted. After that, we are able to directly estimate the GA via a deep regression model (DeepReg). We evaluate DeepGA on a large dataset, including 10, 413 ultrasound images from 7113 subjects. It achieves superior performance over the traditional measurement approach, with a mean absolute estimation error (MAE) of 5 days. Our DeepGA model is a novel automatic solution on the basis of artificial intelligence learning that can help radiologists improve the performance of GA estimation in various clinical scenarios, thereby enhancing the efficiency of prenatal examinations.

JBHI Journal 2023 Journal Article

Estimating Outlier-Immunized Common Harmonic Waves for Brain Network Analyses on the Stiefel Manifold

• Hongmin Cai
• Huan Liu
• Defu Yang
• Guorong Wu
• Bin Hu
• Jiazhou Chen

Since brain network organization is essentially governed by the harmonic waves derived from the Eigen-system of the underlying Laplacian matrix, discovering the harmonic-based alterations provides a new window to understand the pathogenic mechanism of Alzheimer's disease (AD) in a unified reference space. However, current reference (common harmonic waves) estimation studies over the individual harmonic waves are often sensitive to outliers, which are obtained by averaging the heterogenous individual brain networks. To address this challenge, we propose a novel manifold learning approach to identify a set of outlier-immunized common harmonic waves. The backbone of our framework is calculating the geometric median of all individual harmonic waves on the Stiefel manifold, instead of Fréchet mean, thus improving the robustness of learned common harmonic waves to the outliers. A manifold optimization scheme with theoretically guaranteed convergence is tailored to solve our method. The experimental results on synthetic data and real data demonstrate that the common harmonic waves learned by our approach are not only more robust to the outliers than the state-of-the-art methods, but also provide a putative imaging biomarker to predict the early stage of AD.

NeurIPS Conference 2023 Conference Paper

Generalized Information-theoretic Multi-view Clustering

• Weitian Huang
• Sirui Yang
• Hongmin Cai

In an era of more diverse data modalities, multi-view clustering has become a fundamental tool for comprehensive data analysis and exploration. However, existing multi-view unsupervised learning methods often rely on strict assumptions on semantic consistency among samples. In this paper, we reformulate the multi-view clustering problem from an information-theoretic perspective and propose a general theoretical model. In particular, we define three desiderata under multi-view unsupervised learning in terms of mutual information, namely, comprehensiveness, concentration, and cross-diversity. The multi-view variational lower bound is then obtained by approximating the samples' high-dimensional mutual information. The Kullback–Leibler divergence is utilized to deduce sample assignments. Ultimately the information-based multi-view clustering model leverages deep neural networks and Stochastic Gradient Variational Bayes to achieve representation learning and clustering simultaneously. Extensive experiments on both synthetic and real datasets with wide types demonstrate that the proposed method exhibits a more stable and superior clustering performance than state-of-the-art algorithms.

IJCAI Conference 2023 Conference Paper

Invertible Residual Neural Networks with Conditional Injector and Interpolator for Point Cloud Upsampling

• Aihua Mao
• Yaqi Duan
• Yu-Hui Wen
• Zihui Du
• Hongmin Cai
• Yong-Jin Liu

Point clouds obtained by LiDAR and other sensors are usually sparse and irregular. Low-quality point clouds have serious influence on the final performance of downstream tasks. Recently, a point cloud upsampling network with normalizing flows has been proposed to address this problem. However, the network heavily relies on designing specialized architectures to achieve invertibility. In this paper, we propose a novel invertible residual neural network for point cloud upsampling, called PU-INN, which allows unconstrained architectures to learn more expressive feature transformations. Then, we propose a conditional injector to improve nonlinear transformation ability of the neural network while guaranteeing invertibility. Furthermore, a lightweight interpolator is proposed based on semantic similarity distance in the latent space, which can intuitively reflect the interpolation changes in Euclidean space. Qualitative and quantitative results show that our method outperforms the state-of-the-art works in terms of distribution uniformity, proximity-to-surface accuracy, 3D reconstruction quality, and computation efficiency.

JBHI Journal 2022 Journal Article

Identify Multiple Gene-Drug Common Modules via Constrained Graph Matching

• Jiazhou Chen
• Jie Huang
• Yi Liao
• Lei Zhu
• Hongmin Cai

Identifying gene-drug interactions is vital to understanding biological mechanisms and achieving precise drug repurposing. High-throughput technologies produce a large amount of pharmacological and genomic data, providing an opportunity to explore the associations between oncogenic genes and therapeutic drugs. However, most studies only focus on “one-to-one” or “one-to-many” interactions, ignoring the multivariate patterns between genes and drugs. In this article, a high-order graph matching model with hypergraph constraints is proposed to discover the gene-drug common regulatory modules. Moreover, the prior knowledge is formulated into hypergraph constraints to reveal their multiple correspondences, penalizing the tensor matching process. The experimental results on the synthetic data demonstrate the proposed model is robust to noise contamination and outlier corruption, achieving a better performance than four state-of-the-art methods. We then evaluate the statistical power of our proposed method on the pharmacogenomics data. Our identified gene-drug common modules not only show significantly enriched pathways associated with cancer but also manifest the highly close gene-drug interactions.

AAAI Conference 2021 Conference Paper

Savable but Lost Lives when ICU Is Overloaded: a Model from 733 Patients in Epicenter Wuhan, China

• Tingting Dan
• Yang Li
• Ziwei Zhu
• Xijie Chen
• Wuxiu Quan
• Yu Hu
• Guihua Tao
• Lei Zhu

Coronavirus Disease 2019 (COVID-19) causes a sudden turnover to bad at some checkpoints and thus needs the intervention of intensive care unit (ICU). This resulted in urgent and large needs of ICUs posed great risks to the medical system. Estimating the mortality of critical in-patients who were not admitted into the ICU will be valuable to optimize the management and assignment of ICU. Retrospective, 733 in-patients diagnosed with COVID-19 at a local hospital (Wuhan, China), as of March 18, 2020. Demographic, clinical and laboratory results were collected and analyzed using machine learning to build a predictive model. Considering the shortage of ICU beds at the beginning of disease emergence, we defined the mortality for those patients who were predicted to be in needing ICU care yet they did not as Missing-ICU (MI)-mortality. To estimate MI-mortality, a prognostic classification model was built to identify the in-patients who may need ICU care. Its predictive accuracy was 0. 8288, with an AUC of 0. 9119. On our cohort of 733 patients, 25 in-patients who have been predicted by our model that they should need ICU, yet they did not enter ICU due to lack of shorting ICU wards. Our analysis had shown that the MI-mortality is 41%, yet the mortality of ICU is 32%, implying that enough bed of ICU in treating patients in critical conditions.

YNIMG Journal 2006 Journal Article

Repulsive force based snake model to segment and track neuronal axons in 3D microscopy image stacks

• Hongmin Cai
• Xiaoyin Xu
• Ju Lu
• Jeff W. Lichtman
• S.P. Yung
• Stephen T.C. Wong

The branching patterns of axons and dendrites are fundamental structural properties that affect the synaptic connectivity of axons. Although today three-dimensional images of fluorescently labeled processes can be obtained to study axonal branching, there are no robust methods of tracing individual axons. This paper describes a repulsive force based snake model to segment and track axonal profiles in 3D images. This new method segments all the axonal profiles in a 2D image and then uses the results obtained from that image as prior information to help segment the adjacent 2D image. In this way, the segmentation successfully connects axonal profiles over hundreds of images in a 3D image stack. Individual axons can then be extracted based on the segmentation results. The utility and performance of the method are demonstrated using 3D axonal images obtained from transgenic mice that express fluorescent protein.