Yan Chen
Research Subjects
Yan Chen has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
Junhao Chen
Department of Radiology, Jiangmen Central Hospital, Jiangmen, Guangdong, China
2 shared publications- Kuncai Xu
School of Intelligent Engineering, Guiyang Institute of Information Science and Technology, Guiyang, Guizhou, China
2 shared publications Ning Zhang
School of Intelligent Engineering, Guiyang Institute of Information Science and Technology, Guiyang, Guizhou, China
2 shared publications- Deke Wu
Department of Public Teaching, Guiyang Institute of Information Science and Technology, Guiyang, China
1 shared publication - Kuncai Xu
Affiliation College of Intelligent Engineering, Guiyang Institute of Information Science and Technology, Guiyang, Guizhou, PR China
1 shared publication Maorong Li
Department of Pain Medicine, The First Affiliated Hospital of Guizhou University of Traditional Chinese Medicine, Guiyang, Guizhou, China
1 shared publication
Latest Publications
- Open Access
- 14 July 2026
- Published by PeerJ in PeerJ Computer Science
- Vol. 12, e4026
MCCDNet: multi-level collaborative contrast decoupling medical image segmentation network
The aim of this study is to address the problem of low segmentation accuracy in medical image analysis caused by blurred tumour boundaries. We propose a Multi-level Collaborative Contrast Decoupling Network (MCCDNet) for medical image segmentation (MIS). Initially, a contrast consistency enhancement network is constructed to extract complex texture and detail features of tumour lesion boundaries. This ensures that the extracted features can adapt to complex environments. Secondly, we inject the prior knowledge into the multi-level contrast decoupling network to guide further training, enabling the model to capture the key features of the lesion area more effectively. Then, local-global attention module (LGAM) and gated bottleneck convolution (GBC) are used to further refine feature extraction, enhance the model’s focus on key regions, and reduce the uncertainty in the segmentation process. Ultimately, we propose using contrast feature-guided aggregation (CFGA) technology to fuse and enhance multi-level, refined features. This will improve the accuracy of the model in segmenting lesion regions within complex medical images. The experimental results demonstrate that MCCDNet achieves DSC scores of 96.09, 92.08, and 89.29 on the PROMISE12, ISIC2018, and Glas medical image datasets, respectively. It can provide auxiliary reference for precise quantification of clinical lesions and treatment planning.
- 14 July 2026
- Published by PeerJ in PeerJ Computer Science
- Vol. 12, e4026
Open AccessMCCDNet: multi-level collaborative contrast decoupling medical image segmentation network
The aim of this study is to address the problem of low segmentation accuracy in medical image analysis caused by blurred tumour boundaries. We propose a Multi-level Collaborative Contrast Decoupling Network (MCCDNet) for medical image segmentation (MIS). Initially, a contrast consistency enhancement network is constructed to extract complex texture and detail features of tumour lesion boundaries. This ensures that the extracted features can adapt to complex environments. Secondly, we inject the prior knowledge into the multi-level contrast decoupling network to guide further training, enabling the model to capture the key features of the lesion area more effectively. Then, local-global attention module (LGAM) and gated bottleneck convolution (GBC) are used to further refine feature extraction, enhance the model’s focus on key regions, and reduce the uncertainty in the segmentation process. Ultimately, we propose using contrast feature-guided aggregation (CFGA) technology to fuse and enhance multi-level, refined features. This will improve the accuracy of the model in segmenting lesion regions within complex medical images. The experimental results demonstrate that MCCDNet achieves DSC scores of 96.09, 92.08, and 89.29 on the PROMISE12, ISIC2018, and Glas medical image datasets, respectively. It can provide auxiliary reference for precise quantification of clinical lesions and treatment planning.
- Open Access
- 23 March 2026
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 21 (3) , e0339719
LM-UNet: Lightweight Mamba-UNet Prostate MRI image segmentation network
Accurate segmentation of lesions in prostate magnetic resonance images (MRI) is important for assessing patient health and personalized treatment in the clinic. However, the traditional UNet segmentation network has low segmentation accuracy because of the fuzzy boundary and low contrast. Therefore, we propose a Lightweight Mamba-UNet (LM-UNet) prostate MRI image segmentation method. Initially, the encoder-decoder backbone structure consists of parallel vision mamba (PV-Mamba) and efficient multi-scale attention (EMA). The number of model parameters is reduced by constructing PV-Mamba while extracting the correlation between features over long distances. The EMA is then used to learn different spatial features in groups and construct cross-spatial information aggregation methods for richer feature aggregation. Subsequently, we construct the edge feature extraction (EFE) and the edge feature fusion (EFF) to achieve different levels of feature fusion in the encoder. Ultimately, we suggest a multi-stage and multi-level skip connections (MMSC) to achieve multi-level fusion between the encoder and decoder, there reducing semantic discrepancies between contextual features and improving segmentation accuracy. Experimental results demonstrate that on the PROMISE12 dataset, LM-UNet outperforms seven comparative segmentation methods in terms of parameter count, computational memory requirements, and precise segmentation of lesion margins.
- 23 March 2026
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 21 (3) , e0339719
Open AccessLM-UNet: Lightweight Mamba-UNet Prostate MRI image segmentation network
Accurate segmentation of lesions in prostate magnetic resonance images (MRI) is important for assessing patient health and personalized treatment in the clinic. However, the traditional UNet segmentation network has low segmentation accuracy because of the fuzzy boundary and low contrast. Therefore, we propose a Lightweight Mamba-UNet (LM-UNet) prostate MRI image segmentation method. Initially, the encoder-decoder backbone structure consists of parallel vision mamba (PV-Mamba) and efficient multi-scale attention (EMA). The number of model parameters is reduced by constructing PV-Mamba while extracting the correlation between features over long distances. The EMA is then used to learn different spatial features in groups and construct cross-spatial information aggregation methods for richer feature aggregation. Subsequently, we construct the edge feature extraction (EFE) and the edge feature fusion (EFF) to achieve different levels of feature fusion in the encoder. Ultimately, we suggest a multi-stage and multi-level skip connections (MMSC) to achieve multi-level fusion between the encoder and decoder, there reducing semantic discrepancies between contextual features and improving segmentation accuracy. Experimental results demonstrate that on the PROMISE12 dataset, LM-UNet outperforms seven comparative segmentation methods in terms of parameter count, computational memory requirements, and precise segmentation of lesion margins.
- Open Access
- 2 March 2026
- Published by Springer Nature in Journal of King Saud University Computer and Information Sciences
- Vol. 38 (5)
FCSD-Net: A fourier convolution and state-space duality network for multi-modal medical image fusion
In the field of multi-modal medical image fusion (MMIF), combining images from different sources generates a more comprehensive, information-rich fused image. The aim is to improve the quality of clinical diagnostic assistance by providing a more reliable image. To address the aforementioned challenges, we proposed a fourier convolution and state-space duality network (FCSD-Net) for multi-modal medical image fusion. Initially, the image is transformed from the spatial domain to the frequency domain using fourier convolution. This simplifies convolution calculations and effectively extracts image texture and structure, and spatial domain and frequency domain features. Then, we suggested the hidden state interweaver-based state space duality (HSI-SSD), which maps input features onto the hidden state space to reveal task-relevant latent state features. This HSI-SSD extracts deep semantic information while reducing computational complexity. Subsequently, we put forward a dual attention feature discrimination enhancement module (DFEM) that combines dynamic and spatial attention weights. The DFEM improves the ability to discriminate between features and highlights key regions. Ultimately, we advanced the dual-domain fusion mechanism (DDFM), which enhances the unique functional and textural features of both the spatial and frequency domains, while also achieving cross-modal deep fusion. The experimental results on the PET-MRI dataset demonstrate that FCSD-Net achieves AG, SF, SD, Qabf, and $$\varvec{SSIM_{A,B/F}}$$ values of 11.4450, 35.1403, 88.5587, 0.7166, and 1.3386 respectively. FCSD-Net can produce fused images with more comprehensive detail and richer content. This providing more reliable and precise support for clinical decision-making.- 2 March 2026
- Published by Springer Nature in Journal of King Saud University Computer and Information Sciences
- Vol. 38 (5)
Open AccessFCSD-Net: A fourier convolution and state-space duality network for multi-modal medical image fusion
In the field of multi-modal medical image fusion (MMIF), combining images from different sources generates a more comprehensive, information-rich fused image. The aim is to improve the quality of clinical diagnostic assistance by providing a more reliable image. To address the aforementioned challenges, we proposed a fourier convolution and state-space duality network (FCSD-Net) for multi-modal medical image fusion. Initially, the image is transformed from the spatial domain to the frequency domain using fourier convolution. This simplifies convolution calculations and effectively extracts image texture and structure, and spatial domain and frequency domain features. Then, we suggested the hidden state interweaver-based state space duality (HSI-SSD), which maps input features onto the hidden state space to reveal task-relevant latent state features. This HSI-SSD extracts deep semantic information while reducing computational complexity. Subsequently, we put forward a dual attention feature discrimination enhancement module (DFEM) that combines dynamic and spatial attention weights. The DFEM improves the ability to discriminate between features and highlights key regions. Ultimately, we advanced the dual-domain fusion mechanism (DDFM), which enhances the unique functional and textural features of both the spatial and frequency domains, while also achieving cross-modal deep fusion. The experimental results on the PET-MRI dataset demonstrate that FCSD-Net achieves AG, SF, SD, Qabf, and $$\varvec{SSIM_{A,B/F}}$$ values of 11.4450, 35.1403, 88.5587, 0.7166, and 1.3386 respectively. FCSD-Net can produce fused images with more comprehensive detail and richer content. This providing more reliable and precise support for clinical decision-making.