Shuang Liang
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
Shuang Liang has a total of 31 co-authors. Here are the top co-authors by number of shared publications:
- Yu Gu
School of Biomedical Engineering, Capital Medical University, Beijing, China
11 shared publications Qiao Chang
Department of Orthodontics, Beijing Stomatological Hospital, Capital Medical University, Beijing, China
2 shared publicationsShaofeng Wang
Department of Orthodontics, Beijing Stomatological Hospital, Capital Medical University, Beijing, China
2 shared publications- Xianju Xie
Department of Orthodontics, Capital Medical University School of Stomatology, Beijing, China
2 shared publications - Yajie Wang
Department of Engineering Physics, Tsinghua University, Beijing, China
2 shared publications Yu Gu
2 shared publications
Latest Publications
- 19 March 2025
- Published by Elsevier in Journal of Biomedical Informatics
- Vol. 165, 104811
A Language-Guided Progressive Fusion Network with semantic density alignment for Medical Visual Question Answering
- 19 March 2025
- Published by Elsevier in Journal of Biomedical Informatics
- Vol. 165, 104811
A Language-Guided Progressive Fusion Network with semantic density alignment for Medical Visual Question Answering
- Open Access
- 4 February 2025
- Published by Springer Nature in BioMedical Engineering OnLine
- Vol. 24 (1) , 1-16
Automated orthodontic diagnosis via self-supervised learning and multi-attribute classification using lateral cephalograms
Malocclusion, characterized by dental misalignment and improper occlusal relationships, significantly impacts oral health and daily functioning, with a global prevalence of 56%. Lateral cephalogram is a crucial diagnostic tool in orthodontic treatment, providing insights into various structural characteristics. This study introduces a pre-training approach using multi-center lateral cephalograms for self-supervised learning, aimed at improving model generalization across diverse clinical data domains. Additionally, a multi-attribute classification network is proposed, leveraging attribute correlations to optimize parameters and enhance classification performance. Comprehensive evaluation on both public and clinical datasets showcases the superiority of the proposed framework, achieving an impressive average accuracy of 90.02%. The developed Self-supervised Pre-training and Multi-Attribute (SPMA) network achieves a best match ratio (MR) score of 71.38% and a low Hamming loss (HL) of 0.0425%, demonstrating its efficacy in orthodontic diagnosis from lateral cephalograms. This work contributes significantly to advancing automated diagnostic tools in orthodontics, addressing the critical need for accurate and efficient malocclusion diagnosis. The outcomes not only improve the efficiency and accuracy of diagnosis, but also have the potential to reduce healthcare costs associated with orthodontic treatments.
- 4 February 2025
- Published by Springer Nature in BioMedical Engineering OnLine
- Vol. 24 (1) , 1-16
Open AccessAutomated orthodontic diagnosis via self-supervised learning and multi-attribute classification using lateral cephalograms
Malocclusion, characterized by dental misalignment and improper occlusal relationships, significantly impacts oral health and daily functioning, with a global prevalence of 56%. Lateral cephalogram is a crucial diagnostic tool in orthodontic treatment, providing insights into various structural characteristics. This study introduces a pre-training approach using multi-center lateral cephalograms for self-supervised learning, aimed at improving model generalization across diverse clinical data domains. Additionally, a multi-attribute classification network is proposed, leveraging attribute correlations to optimize parameters and enhance classification performance. Comprehensive evaluation on both public and clinical datasets showcases the superiority of the proposed framework, achieving an impressive average accuracy of 90.02%. The developed Self-supervised Pre-training and Multi-Attribute (SPMA) network achieves a best match ratio (MR) score of 71.38% and a low Hamming loss (HL) of 0.0425%, demonstrating its efficacy in orthodontic diagnosis from lateral cephalograms. This work contributes significantly to advancing automated diagnostic tools in orthodontics, addressing the critical need for accurate and efficient malocclusion diagnosis. The outcomes not only improve the efficiency and accuracy of diagnosis, but also have the potential to reduce healthcare costs associated with orthodontic treatments.
- Open Access
A Coarse-to-Fine Feature Aggregation Neural Network with a Boundary-Aware Module for Accurate Food Recognition
Food recognition from images is crucial for dietary management, enabling applications like automated meal tracking and personalized nutrition planning. However, challenges such as background noise disrupting intra-class consistency, inter-class distinction, and domain shifts due to variations in capture angles, lighting, and image resolution persist. This study proposes a multi-stage convolutional neural network-based framework incorporating a boundary-aware module (BAM) for boundary region perception, deformable ROI pooling (DRP) for spatial feature refinement, a transformer encoder for capturing global contextual relationships, and a NetRVLAD module for robust feature aggregation. The framework achieved state-of-the-art performance on three benchmark datasets, with Top-1 accuracies of 99.80% on the Food-5k dataset, 99.17% on the Food-101 dataset, and 85.87% on the Food-2k dataset, significantly outperforming existing methods. This framework holds promise as a foundational tool for intelligent dietary management, offering robust and accurate solutions for real-world applications.
Open AccessA Coarse-to-Fine Feature Aggregation Neural Network with a Boundary-Aware Module for Accurate Food Recognition
Food recognition from images is crucial for dietary management, enabling applications like automated meal tracking and personalized nutrition planning. However, challenges such as background noise disrupting intra-class consistency, inter-class distinction, and domain shifts due to variations in capture angles, lighting, and image resolution persist. This study proposes a multi-stage convolutional neural network-based framework incorporating a boundary-aware module (BAM) for boundary region perception, deformable ROI pooling (DRP) for spatial feature refinement, a transformer encoder for capturing global contextual relationships, and a NetRVLAD module for robust feature aggregation. The framework achieved state-of-the-art performance on three benchmark datasets, with Top-1 accuracies of 99.80% on the Food-5k dataset, 99.17% on the Food-101 dataset, and 85.87% on the Food-2k dataset, significantly outperforming existing methods. This framework holds promise as a foundational tool for intelligent dietary management, offering robust and accurate solutions for real-world applications.
- Open Access
Improving Circulating Tumor Cell Detection Using Image Synthesis and Transformer Models in Cancer Diagnostics
Cancer is the second leading cause of death, significantly threatening human health. Effective treatment options are often lacking in advanced stages, making early diagnosis crucial for reducing mortality rates. Circulating tumor cells (CTCs) are a promising biomarker for early detection; however, their automatic detection is challenging due to their heterogeneous size and shape, as well as their scarcity in blood. This study proposes a data generation method using the Segment Anything Model (SAM) combined with a copy–paste strategy. We develop a detection network based on the Swin Transformer, featuring a backbone network, scale adapter module, shape adapter module, and detection head, which enhances CTC localization and identification in images. To effectively utilize both generated and real data, we introduce an improved loss function that includes a regularization term to ensure consistency across different data distributions. Our model demonstrates exceptional performance across five evaluation metrics: accuracy (0.9960), recall (0.9961), precision (0.9804), specificity (0.9975), and mean average precision (mAP) of 0.9400 at an Intersection over Union (IoU) threshold of 0.5. These results are achieved on a dataset generated by mixing both public and local data, highlighting the robustness and generalizability of the proposed approach. This framework surpasses state-of-the-art models (ADCTC, DiffusionDet, CO-DETR, and DDQ), providing a vital tool for early cancer diagnosis, treatment planning, and prognostic assessment, ultimately enhancing human health and well-being.
Open AccessImproving Circulating Tumor Cell Detection Using Image Synthesis and Transformer Models in Cancer Diagnostics
Cancer is the second leading cause of death, significantly threatening human health. Effective treatment options are often lacking in advanced stages, making early diagnosis crucial for reducing mortality rates. Circulating tumor cells (CTCs) are a promising biomarker for early detection; however, their automatic detection is challenging due to their heterogeneous size and shape, as well as their scarcity in blood. This study proposes a data generation method using the Segment Anything Model (SAM) combined with a copy–paste strategy. We develop a detection network based on the Swin Transformer, featuring a backbone network, scale adapter module, shape adapter module, and detection head, which enhances CTC localization and identification in images. To effectively utilize both generated and real data, we introduce an improved loss function that includes a regularization term to ensure consistency across different data distributions. Our model demonstrates exceptional performance across five evaluation metrics: accuracy (0.9960), recall (0.9961), precision (0.9804), specificity (0.9975), and mean average precision (mAP) of 0.9400 at an Intersection over Union (IoU) threshold of 0.5. These results are achieved on a dataset generated by mixing both public and local data, highlighting the robustness and generalizability of the proposed approach. This framework surpasses state-of-the-art models (ADCTC, DiffusionDet, CO-DETR, and DDQ), providing a vital tool for early cancer diagnosis, treatment planning, and prognostic assessment, ultimately enhancing human health and well-being.
- Open Access
- 2 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 217-224
A Hybrid Foreground Modeling Network for Multiple Fundus Lesion Segmentation Under Class Imbalance
Diabetic retinopathy lesion segmentation in fundus images is challenging, due to the existence of the scale variation, similar lesion appearances, and foreground-background imbalance. The previous semantic segmentation approaches primarily address scale change without sufficiently tackling the other two specific challenges of fundus image segmentation. To solve the aforementioned problems, this paper presents a novel approach, named the HFM-Seg network, aimed to enhance lesion segmentation using a hybrid foreground modeling network. It consists the relation-based and the optimization-based foreground modeling. The relation-based part of the HFM-Seg network uses the cene-context enhancement module to learn the foreground-background relationship to associate the foreground-related multi-scale context, so as to enhance the discrimination of foreground features. And the relation-based part of the HFM-Seg network adopts the adaptive foreground enhancement optimization, which pays attention to both foreground samples and background difficult samples in the training process, achieving balanced optimization. The mAUC values of HFM-Seg on the IDRiD, DDR, and E-optha datasets surpass the average of six other advanced methods by 7.76%, 15.09%, and 13.51% respectively, demonstrating HFM-Seg's effectiveness.
- 2 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 217-224
Open AccessA Hybrid Foreground Modeling Network for Multiple Fundus Lesion Segmentation Under Class Imbalance
Diabetic retinopathy lesion segmentation in fundus images is challenging, due to the existence of the scale variation, similar lesion appearances, and foreground-background imbalance. The previous semantic segmentation approaches primarily address scale change without sufficiently tackling the other two specific challenges of fundus image segmentation. To solve the aforementioned problems, this paper presents a novel approach, named the HFM-Seg network, aimed to enhance lesion segmentation using a hybrid foreground modeling network. It consists the relation-based and the optimization-based foreground modeling. The relation-based part of the HFM-Seg network uses the cene-context enhancement module to learn the foreground-background relationship to associate the foreground-related multi-scale context, so as to enhance the discrimination of foreground features. And the relation-based part of the HFM-Seg network adopts the adaptive foreground enhancement optimization, which pays attention to both foreground samples and background difficult samples in the training process, achieving balanced optimization. The mAUC values of HFM-Seg on the IDRiD, DDR, and E-optha datasets surpass the average of six other advanced methods by 7.76%, 15.09%, and 13.51% respectively, demonstrating HFM-Seg's effectiveness.
- 29 February 2024
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 35 (5) , 055703
Fusion Analyser: a novel measurement method and software tool for dental model analysis in orthodontics
Accurate analysis of dental models is crucial for diagnosing and treating malocclusion successfully. Traditional measurement techniques, however, often fail to accurately measure teeth in crowded conditions. In this study, we introduce Fusion Analyser (FA), an innovative software tool capable of calculating key dental indicators, thereby advancing the scope of digital dental model analysis. Our unique FA framework employs segmentation discs on teeth to enhance the accuracy of the anterior Bolton’s ratio, outperforming the widely-used 3Shape software when dealing with crowded teeth. The FA framework exhibits performance commensurate with the established gold standard, as evidenced by high scores across a range of evaluation metrics, thereby validating its accuracy, stability, and efficiency. The potential of the FA framework as an auxiliary tool for dentists, particularly in diagnosing and strategising treatment plans for malocclusion, is underscored.
- 29 February 2024
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 35 (5) , 055703
Fusion Analyser: a novel measurement method and software tool for dental model analysis in orthodontics
Accurate analysis of dental models is crucial for diagnosing and treating malocclusion successfully. Traditional measurement techniques, however, often fail to accurately measure teeth in crowded conditions. In this study, we introduce Fusion Analyser (FA), an innovative software tool capable of calculating key dental indicators, thereby advancing the scope of digital dental model analysis. Our unique FA framework employs segmentation discs on teeth to enhance the accuracy of the anterior Bolton’s ratio, outperforming the widely-used 3Shape software when dealing with crowded teeth. The FA framework exhibits performance commensurate with the established gold standard, as evidenced by high scores across a range of evaluation metrics, thereby validating its accuracy, stability, and efficiency. The potential of the FA framework as an auxiliary tool for dentists, particularly in diagnosing and strategising treatment plans for malocclusion, is underscored.
- Open Access
- 26 February 2024
- Published by MDPI AG in Diagnostics
- Vol. 14 (5) , 497
STSN-Net: Simultaneous Tooth Segmentation and Numbering Method in Crowded Environments with Deep Learning
Accurate tooth segmentation and numbering are the cornerstones of efficient automatic dental diagnosis and treatment. In this paper, a multitask learning architecture has been proposed for accurate tooth segmentation and numbering in panoramic X-ray images. A graph convolution network was applied for the automatic annotation of the target region, a modified convolutional neural network-based detection subnetwork (DSN) was used for tooth recognition and boundary regression, and an effective region segmentation subnetwork (RSSN) was used for region segmentation. The features extracted using RSSN and DSN were fused to optimize the quality of boundary regression, which provided impressive results for multiple evaluation metrics. Specifically, the proposed framework achieved a top F1 score of 0.9849, a top Dice metric score of 0.9629, and an mAP (IOU = 0.5) score of 0.9810. This framework holds great promise for enhancing the clinical efficiency of dentists in tooth segmentation and numbering tasks.
- 26 February 2024
- Published by MDPI AG in Diagnostics
- Vol. 14 (5) , 497
Open AccessSTSN-Net: Simultaneous Tooth Segmentation and Numbering Method in Crowded Environments with Deep Learning
Accurate tooth segmentation and numbering are the cornerstones of efficient automatic dental diagnosis and treatment. In this paper, a multitask learning architecture has been proposed for accurate tooth segmentation and numbering in panoramic X-ray images. A graph convolution network was applied for the automatic annotation of the target region, a modified convolutional neural network-based detection subnetwork (DSN) was used for tooth recognition and boundary regression, and an effective region segmentation subnetwork (RSSN) was used for region segmentation. The features extracted using RSSN and DSN were fused to optimize the quality of boundary regression, which provided impressive results for multiple evaluation metrics. Specifically, the proposed framework achieved a top F1 score of 0.9849, a top Dice metric score of 0.9629, and an mAP (IOU = 0.5) score of 0.9810. This framework holds great promise for enhancing the clinical efficiency of dentists in tooth segmentation and numbering tasks.
- 22 December 2022
- Published by Springer Nature in Applied Intelligence
- Vol. 53 (13) , 17061-17073
SRENet: a spatiotemporal relationship-enhanced 2D-CNN-based framework for staging and segmentation of kidney cancer using CT images
- 22 December 2022
- Published by Springer Nature in Applied Intelligence
- Vol. 53 (13) , 17061-17073
SRENet: a spatiotemporal relationship-enhanced 2D-CNN-based framework for staging and segmentation of kidney cancer using CT images
- 1 October 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 508-512
A hybrid and fast deep learning framework for Covid-19 detection via 3D Chest CT Images
In this paper, we present a hybrid deep learning framework named CTNet which combines convolutional neural network (CNN) and transformer together for the detection of COVID-19 via 3D chest CT images. It consists of a CNN feature extractor module with SE attention to extract sufficient features from CT scans, together with a transformer model to model the discriminative features of the 3D CT scans. Compared to previous works, CTNet provides an effective and efficient method to perform COVID-19 diagnosis via 3D CT scans with data resampling strategy. Advanced results on a large and public benchmarks, COV19-CT-DB database, was achieved by the proposed CTNet with a macro F1 score of 88.21% on the validation set, which lead ten percentage over the state-of-the-art baseline approach proposed together with the dataset. Notably, the inference speed of the proposed framework is about ten times faster than that of the typical CNN frameworks which make it more promising in actual applications.
- 1 October 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 508-512
A hybrid and fast deep learning framework for Covid-19 detection via 3D Chest CT Images
In this paper, we present a hybrid deep learning framework named CTNet which combines convolutional neural network (CNN) and transformer together for the detection of COVID-19 via 3D chest CT images. It consists of a CNN feature extractor module with SE attention to extract sufficient features from CT scans, together with a transformer model to model the discriminative features of the 3D CT scans. Compared to previous works, CTNet provides an effective and efficient method to perform COVID-19 diagnosis via 3D CT scans with data resampling strategy. Advanced results on a large and public benchmarks, COV19-CT-DB database, was achieved by the proposed CTNet with a macro F1 score of 88.21% on the validation set, which lead ten percentage over the state-of-the-art baseline approach proposed together with the dataset. Notably, the inference speed of the proposed framework is about ten times faster than that of the typical CNN frameworks which make it more promising in actual applications.