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


Cheng Xue has a total of 54 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Prognostic prediction of head and neck cancer through radiomics: a stacking ensemble approach with machine learning and deep learning models

    by Heylie Yt Wong, Cheng Xue, Fuk-Hay Tang, Cynthia Cy Chan, Victoria Ty Li, Sarah Wy Lee

    Aim: This study aimed to develop and evaluate a stacking ensemble machine learning (SEML) model that integrates deep learning (DL) algorithms to improve the accuracy of prognostic predictions for patients with head and neck squamous cell carcinoma (HNSCC). Methods: A cohort of 215 HNSCC patients’ CT images, featuring gross tumor volume (GTV) and planning target volume (PTV) contours, was analyzed. Radiomics features were extracted and converted into quantitative data. These features were then used to train and compare a novel SEML model against standard DL algorithms to predict patient prognosis. Results: The proposed SEML model demonstrated superior predictive performance compared to the DL model, achieving 93% accuracy, 100% sensitivity, and 83% specificity. Statistical analysis using the chi-square test indicated no substantial difference in prediction performance between features derived from GTV and PTV contours (p > 0.05). Conclusions: The SEML model effectively enhances the prognostic prediction accuracy for HNSCC based on radiomic features. This approach shows significant potential to inform clinical decision-making and support the development of customized treatment strategies for improved patient care.

  • TD-SAM: Temporal and Distance-Guided Adaptations of SAM for Accurate Surgical Instrument Segmentation

    by Cheng Xue, Shiyu Zhao, Danqiong Wang, Cheng Chen, Guanyu Yang, Yang Chen

    Accurate automatic surgical instrument segmentation plays a crucial role in robot-assisted surgery, but analyzing surgical videos remains challenging due to factors such as rapid instrument movements, high inter-category similarity, and frequent object occlusions. Current surgical instrument segmentation models struggle to capture both inter-frame variations and intra-frame details in complex surgical scenarios. The Segment Anything Model (SAM) has shown significant potential in various segmentation tasks. However, it has not fully addressed the unique challenges posed by surgical videos. To tackle these issues, we propose a Temporal and Distance-Guided SAM model (TD-SAM) for accurate surgical instrument segmentation. Specifically, we introduce a dynamic cross-frame attention module that effectively captures temporal information across frames, allowing the model to track the dynamic changes of surgical instruments and their environment, thus improving segmentation accuracy. In addition, we present a distance-guided instance refinement module, which enhances the model's ability to distinguish between similar categories, mitigating the class ambiguity caused by inter-category similarity. Extensive experiments on the EndoVis18 and EndoVis17 datasets show that the proposed TD-SAM model outperforms existing models, achieving state-of-the-art performance without using any prompts.


  • DIER ‐Net: Debiased Learning With Medical Image Noisy Label by Intrinsic and Extrinsic Regularization

    by Cheng Xue

    In medical image analysis, the presence of noisy labels and imbalanced data poses significant challenges to the performance of deep learning models, particularly in critical diagnostic tasks. To address this issue, we propose DIER‐Net, a learning with noisy label framework designed to handle noisy labels in imbalanced medical datasets. Our approach introduces a debiased sample selection technique that effectively filters out noisy labels while preserving important minority class samples. Additionally, we employ intrinsic and extrinsic regularization strategies to enhance the model's robustness by leveraging both clean and noisy data. Our method is evaluated on two widely used medical image datasets: the ISIC melanoma classification and Kaggle histopathologic lymph node classification. The experimental results demonstrate that DIER‐Net consistently outperforms existing state‐of‐the‐art methods, particularly in settings with high levels of label noise, offering a robust solution for real‐world clinical applications where noisy and imbalanced data are common. DIER‐Net provides an effective approach to enhance the reliability of AI systems in medical imaging, contributing to more accurate and trustworthy diagnostic outcomes.

  • LADDA: Latent Diffusion-based Domain-adaptive Feature Disentangling for Unsupervised Multi-modal Medical Image Registration

    by Peng Yuan, Jianmin Dong, Wei Zhao, Fei Lyu, Cheng Xue, Yudong Zhang, Chunfeng Yang, Zhan Wu, Zhiqiang Gao, Tianling Lyu, et al.

    Deformable image registration (DIR) is critical for accurate clinical diagnosis and effective treatment planning. However, patient movement, significant intensity differences, and large breathing deformations hinder accurate anatomical alignment in multi-modal image registration. These factors exacerbate the entanglement of anatomical and modality-specific style information, thereby severely limiting the performance of multi-modal registration. To address this, we propose a novel LAtent Diffusion-based Domain-Adaptive feature disentangling (LADDA) framework for unsupervised multi-modal medical image registration, which explicitly addresses the representation disentanglement. First, LADDA extracts reliable anatomical priors from the Latent Diffusion Model (LDM), facilitating downstream content-style disentangled learning. A Domain-Adaptive Feature Disentangling (DAFD) module is proposed to promote anatomical structure alignment further. This module disentangles image features into content and style information, boosting the network to focus on cross-modal content information. Next, a Neighborhood-Preserving Hashing (NPH) is constructed to further perceive and integrate hierarchical content information through local neighbourhood encoding, thereby maintaining cross-modal structural consistency. Furthermore, a Unilateral-Query-Frozen Attention (UQFA) module is proposed to enhance the coupling between upstream prior and downstream content information. The feature interaction within intra-domain consistent structures improves the fine recovery of detailed textures. The proposed framework is extensively evaluated on large-scale multi-center datasets, demonstrating superior performance across diverse clinical scenarios and strong generalization on out-of-distribution (OOD) data.


  • Open Access

    Prognostic Prediction of Head and Neck Cancer through Radiomics: A Stacking Ensemble Approach with Machine Learning and Deep Learning Machine Learning Models

    by Heylie Yt Wong, Cheng Xue, Fuk-Hay Tang, Cynthia Cy Chan, Victoria Cy Li, Sarah Wy Lee

    Head and neck squamous cell carcinoma (HNSCC) poses a major challenge for global healthcare due to its high rates of mortality and morbidity. While radiotherapy remains a primary treatment option, its effectiveness can vary due to tumor heterogeneity. Advanc-es in artificial intelligence (AI) have enabled the application of radiomics to enhance cancer prognosis predictions. Method: This study proposes a stacking ensemble learning approach combined with deep learning models to predict prognosis in HNSCC patients. We utilized a dataset comprising 215 CT images with contoured Gross Tumor Volume (GTV) and Planning Target Volume (PTV) from HNSCC patients. Radiomics features were extracted and analyzed using a stacking ensemble machine learning (SEML) model, while deep learning machine learning (DLML) models were used to optimize prediction performance. Result: Our results indicated that the SEML model outperformed the DLML model in predicting prognosis outcomes, achieving an accuracy of 93%, sensitivity of 100%, and specificity of 83%. No significant difference was found between PTV and GTV for prediction performance (chi-square test, p > 0.05). Conclusion: This study highlights the effectiveness of the SEML model in improving prognostic accuracy for HNSCC pa-tients, with implications for enhancing clinical decision-making and personalizing treatment strategies.

  • GFA-Net: Global Feature Aggregation Network Based on Contrastive Learning for Breast Lesion Automated Segmentation in Ultrasound Images

    by Tao Wang, Chufeng Jin, Yang Chen, Guangquan Zhou, Rongjun Ge, Cheng Xue, Baike Shi, Tianyi Liu, Jean-Louis Coatrieux, Qianjin Feng

    Accurate automatic segmentation of breast lesions in ultrasound images is a challenging auxiliary diagnostic task for automated deployment in breast cancer screening. In this study, a contrastive learning based global feature aggregation network (GFA-Net) is proposed to reduce false detections and missed detections, thereby providing computer vision assistance for the development of automated equipment for breast screening. The method first utilizes the feature extraction layer to extract multi-scale feature maps from the ultrasound image, and uses atrous spatial pyramid pooling (ASPP) to enhance the receptive field of the feature. In order to better utilize the spatial-channel complementary information between multi-scale features, a global feature aggregation (GFA) module is proposed. This module can effectively utilize shallow features to extract deep features. Fine segmentation of tumor boundaries in ultrasound images is also crucial as it can reveal the edge diagnostic features of benign and malignant tumors. Therefore, a result fine repair (RFR) module is developed to refine the boundaries of segmented lesions. In addition, a contrastive deep supervision (CDS) method based on contrastive learning is designed, which can introduce the loss of contrastive learning into the process of deep supervision, and can use the correlation between different data in the same batch of training to improve the feature extraction ability of each extraction stage of the backbone network. The experimental results show that our GFA-Net has better segmentation performance than other existing advanced methods. The applicability analysis also indicates that our method still has good generalization ability for different medical ultrasound images, and still maintains strong competitiveness compared to advanced methods.


  • GlandSAM: Injecting Morphology Knowledge Into Segment Anything Model for Label-Free Gland Segmentation

    by Qixiang Zhang, Yi Li, Cheng Xue, Haonan Wang, Xiaomeng Li

    This paper presents a label-free gland segmentation, GlandSAM, which achieves comparable performance with supervised methods while no label is required during its training or inference phase. We observe that the Segment Anything model produces sub-optimal results on gland dataset: It either over-segments a gland into many fractions or under-segments the gland regions by confusing many of them with the background, due to the complex morphology of glands and lack of sufficient labels. To address this challenge, our GlandSAM innovatively injects two clues about gland morphology into SAM to guide the segmentation process: (1) Heterogeneity within glands and (2) Similarity with the background . Initially, we leverage the clues to decompose the intricate glands by selectively extracting a proposal for each gland sub-region of heterogeneous appearances. Then, we inject the morphology clues into SAM in a fine-tuning manner with a novel morphology-aware semantic grouping module that explicitly groups the high-level semantics of gland sub-regions. In this way, our GlandSAM could capture comprehensive knowledge about gland morphology, and produce well-delineated and complete segmentation results. Extensive experiments conducted on the GlaS dataset and the CRAG dataset reveal that GlandSAM outperforms state-of-the-art label-free methods by a significant margin. Notably, our GlandSAM even surpasses several fully-supervised methods that require pixel-wise labels for training, which highlights the remarkable performance and potential of GlandSAM in the realm of gland segmentation.


  • Cas-PRED: A Limited-Angle Cone Beam CT Reconstruction Research Based on Prior Information

    by Yuan Gao, Liu Zhang, Wenxue Yu, Shijie Wang, Pinzheng Zhang, Cheng Xue, Chunfeng Yang, Shipeng Xie, Gouenou Coatrieux, Jean-Louis Coatrieux, et al.

    Due to limitations in radiation dose or scanning conditions, there are instances where CT data can only be collected within a restricted scanning angle. However, the use of limited-angle projection data often results in reconstructed images that exhibit significant noise and artifacts, which can significantly impact diagnostic accuracy. Conventional iterative reconstruction algorithms, such as the total variation algorithm, consume considerable time simulating the cone beam forward and backward projection processes. Furthermore, these algorithms often struggle to address the challenges of limited-angle reconstruction due to their inherent limitations. To tackle these challenges and achieve superior reconstruction outcomes, we introduce a novel approach known as the Cascaded Prior-Based Residual Encoder-Decoder Network (Cas-PRED). This method employs a multi-level cascaded network as its core architecture, leveraging full-angle CT reconstruction results as prior information to guide the deep learning network’s training process. Our approach consistently delivers superior results compared to existing reconstruction methods, with impressive performance metrics, including a peak signal-to-noise ratio (PSNR) of 33.69±0.90 dB, a structural similarity index (SSIM) of 0.9557±0.0029, and a root mean squared error (RMSE) of 22.73±1.56 for the result of [0, 150°] reconstruction, highlighting its effectiveness in enhancing image quality and accuracy.


  • Prognostic Prediction of Cancer Based on Radiomics Features of Diagnostic Imaging: The Performance of Machine Learning Strategies

    by Fuk-Hay Tang, Cheng Xue, Maria Yy Law, Chui-Ying Wong, Tze-Hei Cho, Chun-Kit Lai

    Tumor phenotypes can be characterized by radiomics features extracted from images. However, the prediction accuracy is challenged by difficulties such as small sample size and data imbalance. The purpose of the study was to evaluate the performance of machine learning strategies for the prediction of cancer prognosis. A total of 422 patients diagnosed with non-small cell lung carcinoma (NSCLC) were selected from The Cancer Imaging Archive (TCIA). The gross tumor volume (GTV) of each case was delineated from the respective CT images for radiomic features extraction. The samples were divided into 4 groups with survival endpoints of 1 year, 3 years, 5 years, and 7 years. The radiomic image features were analyzed with 6 different machine learning methods: decision tree (DT), boosted tree (BT), random forests (RF), support vector machine (SVM), generalized linear model (GLM), and deep learning artificial neural networks (DL-ANNs) with 70:30 cross-validation. The overall average prediction performance of the BT, RF, DT, SVM, GLM and DL-ANNs was AUC with 0.912, 0.938, 0.793, 0.746, 0.789 and 0.705 respectively. The RF and BT gave the best and second performance in the prediction. The DL-ANN did not show obvious advantage in predicting prognostic outcomes. Deep learning artificial neural networks did not show a significant improvement than traditional machine learning methods such as random forest and boosted trees. On the whole, the accurate outcome prediction using radiomics serves as a supportive reference for formulating treatment strategy for cancer patients.


  • Robust Medical Image Classification From Noisy Labeled Data With Global and Local Representation Guided Co-Training

    by Cheng Xue, Lequan Yu, Pengfei Chen, Qi Dou, Pheng-Ann Heng

    Deep neural networks have achieved remarkable success in a wide variety of natural image and medical image computing tasks. However, these achievements indispensably rely on accurately annotated training data. If encountering some noisy-labeled images, the network training procedure would suffer from difficulties, leading to a sub-optimal classifier. This problem is even more severe in the medical image analysis field, as the annotation quality of medical images heavily relies on the expertise and experience of annotators. In this paper, we propose a novel collaborative training paradigm with global and local representation learning for robust medical image classification from noisy-labeled data to combat the lack of high quality annotated medical data. Specifically, we employ the self-ensemble model with a noisy label filter to efficiently select the clean and noisy samples. Then, the clean samples are trained by a collaborative training strategy to eliminate the disturbance from imperfect labeled samples. Notably, we further design a novel global and local representation learning scheme to implicitly regularize the networks to utilize noisy samples in a self-supervised manner. We evaluated our proposed robust learning strategy on four public medical image classification datasets with three types of label noise, i.e. , random noise, computer-generated label noise, and inter-observer variability noise. Our method outperforms other learning from noisy label methods and we also conducted extensive experiments to analyze each component of our method.