Lei Yang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Lei Yang has a total of 34 co-authors. Here are the top co-authors by number of shared publications:
- Huiqin Jiang
School of Electrical and Information Engineering, Zhengzhou University, Zhengzhou, China
6 shared publications - Ling Ma
Zhengzhou University
6 shared publications - Jianbo Gao
Department of Radiology, The First Affiliated Hospital of Zhengzhou University, Zhengzhou, China
5 shared publications - Kang Li
Graduate School of Computer and Artificial Intelligence, Zhengzhou University, Henan, China
2 shared publications - Suya Han
School of Electrical and Information Engineering, Zhengzhou University, Zhengzhou, China
2 shared publications - Yan Li
Key Laboratory of Grain Information Processing and Control, Ministry of Education Henan University of Technology Zhengzhou China
2 shared publications
Latest Publications
- 1 September 2026
- Published by Springer Nature in Signal, Image and Video Processing
- Vol. 20 (11) , 614
A transformer-assisted pulmonary nodule pre-annotation prototype: detection and transfer-learning validation
- 1 September 2026
- Published by Springer Nature in Signal, Image and Video Processing
- Vol. 20 (11) , 614
A transformer-assisted pulmonary nodule pre-annotation prototype: detection and transfer-learning validation
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. PP, 1-13
An Adaptive Fusion Network for Breast Tumor Grading Based on Graph Structure Learning
Breast cancer is a serious threat to women's life and health. Early grading of breast tumor malignancy is the key for improving patient survival rate. In clinical diagnosis, physicians often utilize multi-modal data information to accurately grade. As a new multi-modal data processing method, graph convolutional network is expected to improve the grading accuracy of breast tumors. To solve the problem that a single graph cannot effectively utilize multi-modal information, this paper proposes a multi-graph adaptive fusion network to perform accurate breast tumor grading by using multi-modal data. Firstly, a multi-modal information extraction network is designed to extract specific modal information and graph structure from multi-modal data, which is composed of graph structure learning layer and graph auto-encoder. Secondly, by calculating the attention scores of different modalities as weight coefficients, the modal aggregation features, and modal aggregation graphs are obtained by fusing in a weighted manner. Finally, a neighborhood adaptive aggregation module is designed to gather multi graphs and optimize the conventional graph convolution aggregation mode by calculating the neighborhood aggregation coefficient. Experimental results on both public and private breast tumor datasets show that the graph learning-based diagnosis model can effectively process multi-modal medical data and improve the accuracy of multi grading for breast tumors.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. PP, 1-13
An Adaptive Fusion Network for Breast Tumor Grading Based on Graph Structure Learning
Breast cancer is a serious threat to women's life and health. Early grading of breast tumor malignancy is the key for improving patient survival rate. In clinical diagnosis, physicians often utilize multi-modal data information to accurately grade. As a new multi-modal data processing method, graph convolutional network is expected to improve the grading accuracy of breast tumors. To solve the problem that a single graph cannot effectively utilize multi-modal information, this paper proposes a multi-graph adaptive fusion network to perform accurate breast tumor grading by using multi-modal data. Firstly, a multi-modal information extraction network is designed to extract specific modal information and graph structure from multi-modal data, which is composed of graph structure learning layer and graph auto-encoder. Secondly, by calculating the attention scores of different modalities as weight coefficients, the modal aggregation features, and modal aggregation graphs are obtained by fusing in a weighted manner. Finally, a neighborhood adaptive aggregation module is designed to gather multi graphs and optimize the conventional graph convolution aggregation mode by calculating the neighborhood aggregation coefficient. Experimental results on both public and private breast tumor datasets show that the graph learning-based diagnosis model can effectively process multi-modal medical data and improve the accuracy of multi grading for breast tumors.
- 11 November 2025
- Published by Springer Nature
- p. 60-72
FGFENet: Fine-Grained Feature Enhancement Network for Brain Tumor Segmentation
- 11 November 2025
- Published by Springer Nature
- p. 60-72
FGFENet: Fine-Grained Feature Enhancement Network for Brain Tumor Segmentation
- Open Access
- 1 November 2025
- Published by IOP Publishing in Journal of Physics: Conference Series
- Vol. 3135 (1)
A Detection Approach for Traumatic Bleeding Based on the Fusion of Deep Learning Networks and Multi Thresholds
Bleeding caused by trauma is a severe threat to human health. In this study, an approach of detecting traumatic bleedings from contrast-enhanced CT images is proposed to provide early diagnosis results for references of medical doctors. The proposed method can be divided into deep learning network-based and adaptive thresholding-based approaches. The deep learning part uses three modified segmentation networks to segment the input CT case based on the 3D U-net and self-attention mechanism. Post-processing based on CT value thresholds is applied to get global features and boundaries. The adaptive threshold part analyzes the CT value of all bleeding areas of the data set. It formulates multiple thresholds to extract possible bleeding areas from input cases. A random forest-based classifier is trained by texture features to decrease the number of false-positive predictions further. Finally, the results of the above two methods are fused by a proposed coincidence detection method. The overall result gets 82.14% sensitivity and 129 average false positives per case.
- 1 November 2025
- Published by IOP Publishing in Journal of Physics: Conference Series
- Vol. 3135 (1)
Open AccessA Detection Approach for Traumatic Bleeding Based on the Fusion of Deep Learning Networks and Multi Thresholds
Bleeding caused by trauma is a severe threat to human health. In this study, an approach of detecting traumatic bleedings from contrast-enhanced CT images is proposed to provide early diagnosis results for references of medical doctors. The proposed method can be divided into deep learning network-based and adaptive thresholding-based approaches. The deep learning part uses three modified segmentation networks to segment the input CT case based on the 3D U-net and self-attention mechanism. Post-processing based on CT value thresholds is applied to get global features and boundaries. The adaptive threshold part analyzes the CT value of all bleeding areas of the data set. It formulates multiple thresholds to extract possible bleeding areas from input cases. A random forest-based classifier is trained by texture features to decrease the number of false-positive predictions further. Finally, the results of the above two methods are fused by a proposed coincidence detection method. The overall result gets 82.14% sensitivity and 129 average false positives per case.
- Open Access
- 13 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 59-65
GLCFANet: Liver Tumor Segmentation with Global and Local Information Cross-Fusion Attention
Liver tumour edges are fuzzy and difficult to distinguish, making it challenging to clearly define the boundary. The traditional deep learning-based liver tumour segmentation method has limited segmentation accuracy due to its inability to comprehensively extract the tumour's feature information. In this paper, we propose a liver tumor segmentation approach based on Global and Local Information Cross-Fusion Attention(GLCFA). Firstly, a bi-feature extractor module(BEM) is designed, which introduces dilated convolution combined with transformer to extract global features. At the same time, a CNN-based gated convolution block is designed to extract local features. Secondly, a bi-feature cross fusion module(BCFM) is proposed to cross-merge global and local information. This module adopts feature grouping and cross-splicing strategies, and then utilizes a channel global and local convolutional attention mechanism to dynamically calculate their feature importance weight coefficients, thereby effectively fusing the tumour features. Finally, a U-shaped segmentation architecture is employed to complete liver tumour segmentation. The proposed method is evaluated on the LITS2017 dataset and the 3DIRCADb-01 dataset, with the Dice coefficients improved by 3.54% and 5.35%, respectively. This proves the superiority of the method in segmenting liver tumours with ambiguous boundaries and is expected to provide a quantitative basis for accurate diagnosis of liver tumours.
- 13 August 2024
- Published by Association for Computing Machinery (ACM)
- p. 59-65
Open AccessGLCFANet: Liver Tumor Segmentation with Global and Local Information Cross-Fusion Attention
Liver tumour edges are fuzzy and difficult to distinguish, making it challenging to clearly define the boundary. The traditional deep learning-based liver tumour segmentation method has limited segmentation accuracy due to its inability to comprehensively extract the tumour's feature information. In this paper, we propose a liver tumor segmentation approach based on Global and Local Information Cross-Fusion Attention(GLCFA). Firstly, a bi-feature extractor module(BEM) is designed, which introduces dilated convolution combined with transformer to extract global features. At the same time, a CNN-based gated convolution block is designed to extract local features. Secondly, a bi-feature cross fusion module(BCFM) is proposed to cross-merge global and local information. This module adopts feature grouping and cross-splicing strategies, and then utilizes a channel global and local convolutional attention mechanism to dynamically calculate their feature importance weight coefficients, thereby effectively fusing the tumour features. Finally, a U-shaped segmentation architecture is employed to complete liver tumour segmentation. The proposed method is evaluated on the LITS2017 dataset and the 3DIRCADb-01 dataset, with the Dice coefficients improved by 3.54% and 5.35%, respectively. This proves the superiority of the method in segmenting liver tumours with ambiguous boundaries and is expected to provide a quantitative basis for accurate diagnosis of liver tumours.
- 5 November 2023
- Published by Springer Nature
- p. 312-325
ABFNet: Attention Bottlenecks Fusion Network for Multimodal Brain Tumor Segmentation
- 5 November 2023
- Published by Springer Nature
- p. 312-325
ABFNet: Attention Bottlenecks Fusion Network for Multimodal Brain Tumor Segmentation
- 26 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 27 (12) , 5994-6005
MSA-GCN: A Multi-information Selection Aggregation Graph Convolutional Network for Breast Tumor Grading
Physicians typically combine multi-modal data to make a graded diagnosis of breast tumors. However, most existing breast tumor grading methods rely solely on image information, resulting in limited accuracy in grading. This paper proposes a Multi-information Selection Aggregation Graph Convolutional Networks (MSA-GCN) for breast tumor grading. Firstly, to fully utilize phenotypic data reflecting the clinical and pathological characteristics of tumors, an automatic combination screening and weight encoder is proposed for phenotypic data, which can construct a population graph with improved structural information. Then, a graph structure is designed through similarity learning to reflect the correlation between patient image features. Finally, a multi-information selection aggregation mechanism is employed in the graph convolution model to extract the effective features of multi-modal data and enhance the classification performance of the model. The proposed method is evaluated on different clinical datasets from the Digital Database for Screening Mammography (DDSM) and INbreast. The average classification accuracies are 90.74% and 85.35%, respectively, surpassing the performance of existing methods. In conclusion, our method effectively fuses image and non-image information, leading to a significant improvement in the accuracy of breast tumor grading.
- 26 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 27 (12) , 5994-6005
MSA-GCN: A Multi-information Selection Aggregation Graph Convolutional Network for Breast Tumor Grading
Physicians typically combine multi-modal data to make a graded diagnosis of breast tumors. However, most existing breast tumor grading methods rely solely on image information, resulting in limited accuracy in grading. This paper proposes a Multi-information Selection Aggregation Graph Convolutional Networks (MSA-GCN) for breast tumor grading. Firstly, to fully utilize phenotypic data reflecting the clinical and pathological characteristics of tumors, an automatic combination screening and weight encoder is proposed for phenotypic data, which can construct a population graph with improved structural information. Then, a graph structure is designed through similarity learning to reflect the correlation between patient image features. Finally, a multi-information selection aggregation mechanism is employed in the graph convolution model to extract the effective features of multi-modal data and enhance the classification performance of the model. The proposed method is evaluated on different clinical datasets from the Digital Database for Screening Mammography (DDSM) and INbreast. The average classification accuracies are 90.74% and 85.35%, respectively, surpassing the performance of existing methods. In conclusion, our method effectively fuses image and non-image information, leading to a significant improvement in the accuracy of breast tumor grading.
- 18 October 2023
- Published by Springer Nature in Signal, Image and Video Processing
- Vol. 18 (1) , 843-852
A 3D prediction model for benign or malignant of pulmonary nodules based on neural architecture search
- 18 October 2023
- Published by Springer Nature in Signal, Image and Video Processing
- Vol. 18 (1) , 843-852
A 3D prediction model for benign or malignant of pulmonary nodules based on neural architecture search