HZHaixing Zhu

Haixing Zhu

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Haixing Zhu has a total of 38 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • A remaining useful life prediction network using spatio-temporal soft threshold unit and adaptive multi-task weight allocation with missing sensor data

    by Weipeng Liu, Jingtian Shuai, Qidi Hu, Yuxiang Hu, Xiaofei Xue, Haixing Zhu, Jixuan Wang

    Accurate remaining useful life (RUL) prediction is essential for predictive maintenance. However, incomplete sensor data disrupts spatio-temporal dependencies and degrades prediction accuracy. The existing two-stage methods generate the amplification of errors by decoupling interpolation and prediction, and the end-to-end models struggle to capture both spatial coupling and temporal degradation continuity. Moreover, fixed task weights further limit adaptability under varying levels of missing data. To overcome these challenges, this paper proposes a multi-task spatio-temporal residual shrinkage network (ST-RSMTNet) combined a spatio-temporal residual shrinkage block (ST-RSBU) with a progressive multi-task learning strategy. The introduction of ST-RSBU adaptively filters degradation-related features across sensor channels and restores interrupted temporal trends via global sequence statistics. The proposed multi-task framework jointly optimizes imputation and prediction within an "imputation-predict" closed loop. Extensive experiments on the C-MAPSS and PHM2010 datasets show that the ST-RSMTNet maintains robustness under high missing rates. It achieves RMSE reductions of 7.29% on FD001 and 10.83% on PHM2010 C1 compared with the second-best approach. This network outperforms most state-of-the-art methods. The results demonstrate that the ST-RSMTNet is an effective solution for industrial prognostics under incomplete sensing condition. This work proposes a novel solution to the data deficiency challenges and provides crucial theoretical support for reliability of complex industrial systems in the future.

  • Nonlinear coupling calculation method based on Kolmogorov-Arnold attention dynamic spatiotemporal graph for remaining useful life prediction

    by Weipeng Liu, Sheng Gao, Qidi Hu, Yuxiang Hu, Haixing Zhu, Xiaofei Xue, Chen Liu, Gao Tianshi, Jixuan Wang

    Remaining useful life (RUL) prediction is a key aspect of prognostics and health management. It is essential for ensuring the reliability and operational safety of mechanical equipment. Graph Attention Networks (GATs) are an emerging approach for RUL prediction through the innovative integration of graph structures and attention mechanisms. However, existing graph structures fail to establish the correlations between different sensors at different timestamps. Furthermore, the attention mechanism of traditional GATs exhibits insufficient capability in capturing complex nonlinear dependencies between nodes. To overcome these limitations, this paper proposes a nonlinear coupling spatiotemporal graph attention network (NCST-GAT) for RUL prediction. First, a spatiotemporal graph structure construction method is designed to integrate sensor nodes across timestamps. Then, a multiplicative Kolmogorov-Arnold attention mechanism is proposed to replace the standard multilayer perceptron in attention scoring functions, significantly enhancing the modeling of nonlinear interactions among nodes and yielding more expressive feature representations to predict RUL. Experimental results on the CMAPSS and PHM2010 datasets demonstrate that the proposed NCST-GAT achieves an average improvement of 6.1% and 15.6% for RMSE and prognostic Score metrics respectively over state-of-the-art methods, exhibiting strong predictive performance and generalization capability.


  • Cross-teaching with dual uncertainty awareness for semi-supervised medical image segmentation

    by Qinglong Xu, Haixing Zhu, Yuan Wang, Zhongjie Shi, Weipeng Liu


  • Open Access

    CMAGIN

    by Weipeng Liu, Bowen Shi, Jixuan Wang, Xiangyu Hu, Xiaofei Xue, Haixing Zhu

    Multivariate time series (MTS) data is widely utilized in industrial manufacturing, equipment maintenance, and health monitoring. However, the high dimensionality, dynamic nature, and heterogeneity characteristics bring significant challenges for modeling. Traditional deep learning algorithms based on sequential modeling struggle to capture the complex structural relationships between different time series variables, making it difficult to uncover interaction patterns and potential dependencies. To address the dynamic and complex dependencies among variables in MTS data and further balance the importance distribution across multiple temporal feature channels, this work proposes a channel-aware multi-scale adaptive graph interaction network (CMAGIN) for MTS forecasting. The proposed framework integrates a dynamic and adaptive graph constructor with local awareness and global attention (DAGC-LAGA) and a channel-wise adaptive center enhancement (CACE) mechanism. The design of DAGC-LAGA captures sparse neighborhood relations through a multi-view local dynamic graph constructor and further leverages a global attention graph enhancer to model semantic correlations. The results effectively display dynamic dependencies among variables. The introduction of the CACE module dynamically enhances key node features by calculating the node importance at the channel level. In addition, applying the centrality-aware attention mechanism improves the sensitivity of the model to crucial temporal patterns. Furthermore, the results are verified via the C-MAPSS dataset for aircraft engine degradation prediction. Experimental results demonstrate that the CMAGIN model outperforms comparative methods in both RMSE and Score metrics, and exhibits robust performance under complex operating conditions and multiple-fault scenarios. Future research could investigate scalable applications of CMAGIN across diverse industrial scenarios to enable field deployment of intelligent operation and maintenance systems.

  • A robust framework with knowledge-guided planning and fiducial-based registration for bronchoscopy navigation system

    by Haixing Zhu, Zhongjie Shi, Wenbo Zhai, Yifei Liu, Yuan Wang, Zimo Bai, Zhanxiang Wang, Rining Wu, Weipeng Liu

    Objective. Bronchoscopy is a valuable minimally invasive examination in clinical practice and is widely used in the diagnosis of suspected peripheral lung lesions. However, this device has difficulty in accessing peripheral areas from lack of inadequate guidance and heavily relies on intraoperative x-ray or computerized tomography scan. Approach. In order to overcome these limitations, we propose a robust navigation framework with knowledge-guided planning and fiducial-based registration for bronchoscopy navigation, which makes three notable contributions that have been experimentally verified to be of practical value. Firstly, we propose a preoperative path-planning algorithm with anatomical prior knowledge to generate a feasible and accurate trajectory. Secondly, a fiducial-based patient-image registration scheme is introduced to align virtual images with the patient’s actual anatomy, building a robust rigid relationship and enabling real-time updates. Thirdly, we establish a position sensing approach to present precise positions and track the movement of the bronchoscope. Main results. Extensive experiments on the 3D-printed airway tree model and in vivo porcine lung are conducted to evaluate comprehensive capabilities. Qualitative and quantitative results manifest that our framework can achieve excellent performance, reaching a success rate of 100% in the path-planning stage, achieving robust registration precision with a fiducial registration error of 0.998 ± 0.074 mm, and obtaining the standard deviation of 0.017 mm, 0.026 mm and 0.013 mm in the tracking stage. Significance. Our results demonstrate the feasibility and effectiveness and further has potential prospects as an auxiliary tool to extend the capabilities of clinical bronchoscopy.


  • Explainable Classification of Benign-Malignant Pulmonary Nodules With Neural Networks and Information Bottleneck

    by Haixing Zhu, Weipeng Liu, Zhifan Gao, Heye Zhang

    Computerized tomography (CT) is a clinically primary technique to differentiate benign-malignant pulmonary nodules for lung cancer diagnosis. Early classification of pulmonary nodules is essential to slow down the degenerative process and reduce mortality. The interactive paradigm assisted by neural networks is considered to be an effective means for early lung cancer screening in large populations. However, some inherent characteristics of pulmonary nodules in high-resolution CT images, e.g., diverse shapes and sparse distribution over the lung fields, have been inducing inaccurate results. On the other hand, most existing methods with neural networks are dissatisfactory from a lack of transparency. In order to overcome these obstacles, a united framework is proposed, including the classification and feature visualization stages, to learn distinctive features and provide visual results. Specifically, a bilateral scheme is employed to synchronously extract and aggregate global-local features in the classification stage, where the global branch is constructed to perceive deep-level features and the local branch is built to focus on the refined details. Furthermore, an encoder is built to generate some features, and a decoder is constructed to simulate decision behavior, followed by the information bottleneck viewpoint to optimize the objective. Extensive experiments are performed to evaluate our framework on two publicly available datasets, namely, 1) the Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI) and 2) the Lung and Colon Histopathological Image Dataset (LC25000). For instance, our framework achieves 92.98% accuracy and presents additional visualizations on the LIDC. The experiment results show that our framework can obtain outstanding performance and is effective to facilitate explainability. It also demonstrates that this united framework is a serviceable tool and further has the scalability to be introduced into clinical research.