Article
Engineering, Electrical & Electronic
Yunjia Huang, Haixia Xu
Summary: The paper proposes a fully convolutional network with attention modules for semantic segmentation, which enhances pixel relevancy through post-processing and skip-layer attention, optimizes the loss function, and improves performance compared to other models.
SIGNAL IMAGE AND VIDEO PROCESSING
(2021)
Article
Engineering, Multidisciplinary
Hongan Li, Jiangwen Fan, Qiaozhi Hua, Xinpeng Li, Zheng Wen, Meng Yang
Summary: This paper proposes a biomedical sensor image segmentation method with improved fully convolutional network, which effectively extracts local spatial and texture information of the images, suppresses background interference, and enhances image features for better segmentation effect and accuracy.
Article
Computer Science, Artificial Intelligence
Qing Liu, Yongsheng Dong, Yuanhua Pei, Lintao Zheng, Lei Zhang
Summary: In this paper, a Long and Short-Range Relevance Context Network is proposed to capture global semantic context and local spatial context information. The network utilizes Long-Range Relevance Context Module and Short-Range Relevance Context Module to improve the accuracy of pixel classification and detailed pixel location. A coding and decoding structure is adopted to enhance the segmentation results, and experiments on multiple datasets validate the effectiveness of the network.
COMPLEX & INTELLIGENT SYSTEMS
(2023)
Article
Computer Science, Information Systems
Weihao Weng, Xin Zhu, Lei Jing, Mianxiong Dong
Summary: This paper introduces a novel architecture, smooth attention branch (SAB), that simplifies the understanding of long-range pixel-pixel dependencies in small-scale biomedical image segmentation. SAB is a modified attention operation that implements a subnetwork using reshaped feature maps rather than directly calculating a softmax value for attention scores. SAB fuses multilayer attentive feature maps to learn visual attention in multilevel features.
Article
Engineering, Biomedical
Jingdong Yang, Jintu Zhu, Hailing Wang, Xin Yang
Summary: This study proposed a Dilated MultiResUNet network to improve end-to-end image segmentation performance based on U-Net, Res2Net, MultiResUNet, Dilated Residual Networks, and Squeeze-and-Excitation Networks. Evaluation on four biomedical datasets showed superior accuracy and generalization performance with significantly fewer parameters compared to the U-Net model.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2021)
Article
Chemistry, Analytical
Chih-Chiang Wei, Tzu-Heng Huang
Summary: The study utilized fully convolutional networks to establish a forecasting model for predicting hourly rainfall during typhoons in Taiwan. By deep learning and image recognition technology, the model effectively improved the accuracy of rainfall forecasting during typhoons in southern Taiwan.
Article
Geography, Physical
Xue Yang, Xiang Fan, Mingjun Peng, Qingfeng Guan, Luliang Tang
Summary: This study proposes a new model, AD-HRNet, for semantic segmentation of remote sensing images. It addresses challenges such as unbalanced category weight, rich context leading to recognition difficulties, and blurred boundaries of multi-scale objects. The model combines HRNet with attention mechanisms and dilated convolution, and achieves improved performance in terms of mIoUs on different datasets.
INTERNATIONAL JOURNAL OF DIGITAL EARTH
(2022)
Article
Computer Science, Artificial Intelligence
Omar M. Saad, Wei Chen, Fangxue Zhang, Liuqing Yang, Xu Zhou, Yangkang Chen
Summary: In this paper, a fully convolutional DenseNet method for automatic salt segmentation is proposed, with a squeeze-and-excitation network used as a self-attention mechanism to extract the important information related to the salt signals. The method demonstrates robust performance when applied to new datasets using transfer learning and a small amount of training data.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Mohammed A. M. Elhassan, Chenxi Huang, Chenhui Yang, Tewodros Legesse Munea
Summary: Efficient and accurate semantic segmentation is crucial in scene understanding for autonomous driving. This paper introduces a computationally efficient network named DSANet, which utilizes a two-branch strategy to address real-time semantic segmentation in urban scenes. The proposed method achieves high segmentation accuracy while improving inference speed through semantic encoding, dual attention modules, and spatial encoding network.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Biology
Zekun Wang, Yanni Zou, Peter X. Liu
Summary: This study introduces an improved U-Net model, HDA-ResUNet, with residual connections, channel attention block, and hybrid dilated attention convolutional layer for medical image segmentation, achieving more accurate results than traditional U-Net with faster convergence speed. Evaluated on four datasets, the proposed model outperforms U-Net in terms of segmentation accuracy and parameters efficiency.
COMPUTERS IN BIOLOGY AND MEDICINE
(2021)
Article
Engineering, Biomedical
Jiangchang Xu, Jiannan Liu, Dingzhong Zhang, Zijie Zhou, Xiaoyi Jiang, Chenping Zhang, Xiaojun Chen
Summary: An automatic mandibular segmentation method based on neural network with DenseASPP and AG modules was proposed to enhance accuracy. Test results showed that the network achieved good segmentation results with high accuracy, which was close to ground truth.
INTERNATIONAL JOURNAL OF COMPUTER ASSISTED RADIOLOGY AND SURGERY
(2021)
Article
Computer Science, Artificial Intelligence
Ahmed Iqbal, Muhammad Sharif, Muhammad Attique Khan, Wasif Nisar, Majed Alhaisoni
Summary: Automatic multimodal image segmentation is a challenging research area in the biomedical field. This study proposes modifications to the classical UNet architecture by adjusting the receptive field. The proposed models achieve improved performance through post-processing schemes and are validated on multiple biomedical datasets.
COGNITIVE COMPUTATION
(2022)
Article
Multidisciplinary Sciences
Naga Raju Gudhe, Hamid Behravan, Mazen Sudah, Hidemi Okuma, Ritva Vanninen, Veli-Matti Kosma, Arto Mannermaa
Summary: In this study, a novel multi-level dilated residual neural network is proposed for biomedical image segmentation. By replacing convolutional blocks of the classical U-Net with multi-level dilated residual blocks and incorporating non-linear multi-level residual blocks into skip connections, enhanced segmentation performance is achieved.
SCIENTIFIC REPORTS
(2021)
Article
Computer Science, Artificial Intelligence
Jun Zhang, Zhiyuan Hua, Kezhou Yan, Kuan Tian, Jianhua Yao, Eryun Liu, Mingxia Liu, Xiao Han
Summary: This paper introduces a weakly-supervised model using joint fully convolutional and graph convolutional networks for automated segmentation of pathology images. By utilizing image-level labels instead of pixel-wise annotations, the segmentation model's performance is improved. Experimental results demonstrate the effectiveness of this method in cancer region segmentation.
MEDICAL IMAGE ANALYSIS
(2021)
Article
Computer Science, Information Systems
Xiaogen Zhou, Xingqing Nie, Zhiqiang Li, Xingtao Lin, Ensheng Xue, Luoyan Wang, Junlin Lan, Gang Chen, Min Du, Tong Tong
Summary: Skin lesions and thyroid cancer are common diseases, and a computer-aided diagnosis system can improve accuracy. This study proposes a novel dual encoder-decoder network (H-Net) for automated segmentation of thyroid nodules and skin lesions. Experimental results show that H-Net outperforms other methods on multiple datasets.
INFORMATION SCIENCES
(2022)
Article
Computer Science, Information Systems
Hongfeng You, Long Yu, Shengwei Tian, Xiang Ma, Yan Xing
Summary: In this paper, a new deep learning algorithm (TIC-Net) is proposed for medical image segmentation. By utilizing multiple structures and feature fusion, the algorithm fully mines the semantic information between features and achieves the best results.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Xinyu Zhang, Long Yu, Shengwei Tian
Summary: In today's social media and lifestyle applications, expressing sentiments through comments or instant barrage is common. Aspect-based sentiment analysis has become a widely-used technology, using public datasets as benchmarks. Current models stack multi-RNNs layers or combine neural networks with pre-trained models. Considering the importance of dependencies between aspect words and sentiment words, a novel model (BGAT) blending BiGRU and RGAT is investigated to learn dependencies information. Extensive experiments on five datasets demonstrate the great capability of the model.
JOURNAL OF INTELLIGENT & FUZZY SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Jing Liu, Shengwei Tian, Long Yu, Jun Long, Tiejun Zhou, Bo Wang
Summary: This paper proposes a multi-modal fusion sarcasm detection model based on the attention mechanism, introducing Vision Transformer (ViT) to extract image features and designing a Double-Layer Bi-Directional Gated Recurrent Unit (D-BiGRU) to extract text features. The features of the two modalities are fused into one feature vector and predicted after attention enhancement. The model presented in this paper achieved significant experimental results on the baseline datasets, with F1-score and accuracy higher by 0.71% and 0.38% respectively compared to the best baseline model.
JOURNAL OF INTELLIGENT & FUZZY SYSTEMS
(2023)
Article
Health Care Sciences & Services
Yongtao Wang, Shengwei Tian, Long Yu, Weidong Wu, Dezhi Zhang, Junwen Wang, Junlong Cheng
Summary: To improve the feature expression ability and segmentation performance of U-Net, we proposed a feature supplement and optimization U-Net (FSOU-Net). The proposed method utilizes shallow feature supplement module and deep feature optimization module to enhance the representation ability of features. Experimental results demonstrate the superiority of the proposed model in medical image segmentation.
TECHNOLOGY AND HEALTH CARE
(2023)
Correction
Computer Science, Information Systems
Hongfeng You, Long Yu, Shengwei Tian, Xiang Ma, Yan Xing
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Engineering, Biomedical
Bin Yu, Long Yu, Shengwei Tian, Weidong Wu, Zhang Dezhi, Xiaojing Kang
Summary: This research proposes a new multi-scale channel attention module (MS-CA), which is applied to an image segmentation model for accurate diagnosis and treatment planning of skin lesions. Experimental results show that the MS-CA model achieves better segmentation results compared to existing methods.
COMPUTER METHODS IN BIOMECHANICS AND BIOMEDICAL ENGINEERING-IMAGING AND VISUALIZATION
(2023)
Article
Engineering, Biomedical
Zhen Peng, Shengwei Tian, Long Yu, Dezhi Zhang, Weidong Wu, Shaofeng Zhou
Summary: Semi-supervised learning is significant in medical imaging tasks, but pseudo-labeling-based methods face two problems in medical image datasets: bias towards the majority class and loss of useful information. To address these issues, we propose FullMatch, an SSL framework that utilizes all unlabeled data. Our method includes adaptive threshold pseudo-labeling (ATPL) that generates pseudo-labels based on the model's learning status and does not discard unlabeled data below the thresholds. We also introduce unreliable sample contrastive loss (USCL) to leverage useful information from low-confidence unlabeled data. Experimental results demonstrate the superiority of our method over state-of-the-art SSL methods.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2023)
Article
Computer Science, Interdisciplinary Applications
Shaofeng Zhou, Shenwei Tian, Long Yu, Weidong Wu, Dezhi Zhang, Zhen Peng, Zhicheng Zhou
Summary: In this article, we propose a semi-supervised learning method called ReFixMatch-LS, which combines consistency regularization and pseudo-labeling. We apply this method to medical image classification and reduce the impact of noisy labels through label smoothing and consistency regularization. By recording high-confidence pseudo-labels generated during training, we effectively increase the number of pseudo-labels and improve the model performance.
MEDICAL & BIOLOGICAL ENGINEERING & COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Fan Wang, Shengwei Tian, Long Yu, Jing Liu, Junwen Wang, Kun Li, Yongtao Wang
Summary: In this study, a multimodal encoding-decoding translation network with a Transformer was proposed to address the impact of individual modal data on sentiment analysis results. The model achieved improved accuracy by using a joint encoding-decoding method, a modality reinforcement cross-attention module, and a dynamic filtering mechanism.
COGNITIVE COMPUTATION
(2023)
Article
Biochemical Research Methods
Jinmiao Song, Shengwei Tian, Long Yu, Qimeng Yang, Yuanxu Wang, Qiguo Dai, Xiaodong Duan
Summary: Studies have shown that IncRNA-miRNA interactions have important effects on gene expression and biological activities. In this research, a new prediction model called ISLMI was proposed, which used information injection and a second order graph convolution network (SOGCN) to enhance the performance of predicting lncRNA-miRNA interactions. The model achieved reliable performance in 5-fold cross-validation and significantly improved the prediction accuracy.
IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Hongfeng You, Long Yu, Shengwei Tian, Weiwei Cai
Summary: We propose a stereo spatial discoupling network (TSDNets) to leverage the multi-dimensional spatial details of medical images, solve the difficulty of establishing effective spatial associations, and extract similar low-level features, resulting in redundancy of information.
COMPLEX & INTELLIGENT SYSTEMS
(2023)
Article
Engineering, Biomedical
Shaofeng Zhou, Shenwei Tian, Long Yu, Weidong Wu, Dezhi Zhang, Zhen Peng, Zhicheng Zhou, Junwen Wang
Summary: Recent research in semi-supervised learning focuses on consistency regularization using data augmentation, while the more general method of pseudolabelling is limited by noisy training. Medical datasets have a long-tail distribution, and combining these limitations, we propose FixMatch-LS and its variant FixMatch-LS-v2 for medical image classification. We introduce label smoothing to adjust the pseudolabel threshold and reduce the influence of noisy pseudolabels, and emphasize the importance of consistency for pseudolabelling to improve pseudolabel quality. The framework is validated on skin lesion diagnoses from the ISIC 2018 and ISIC 2019 challenges, achieving AUCs of 91.63%, 93.70%, 94.46%, and 95.44% on different proportions of labelled data from ISIC 2018.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2023)
Article
Computer Science, Artificial Intelligence
Qianghua Liu, Yu Tian, Tianshu Zhou, Kewei Lyu, Ran Xin, Yong Shang, Ying Liu, Jingjing Ren, Jingsong Li
Summary: This study proposes a few-shot disease diagnosis decision making model based on a model-agnostic meta-learning algorithm (FSDD-MAML). It significantly improves the diagnostic process in primary health care and helps general practitioners diagnose few-shot diseases more accurately.
ARTIFICIAL INTELLIGENCE IN MEDICINE
(2024)
Article
Computer Science, Artificial Intelligence
Balazs Borsos, Corinne G. Allaart, Aart van Halteren
Summary: The study demonstrates the feasibility of predicting functional outcomes for ischemic stroke patients and the usability of multimodal deep learning architectures for this purpose.
ARTIFICIAL INTELLIGENCE IN MEDICINE
(2024)
Article
Computer Science, Artificial Intelligence
Abdelmoniem Helmy, Radwa Nassar, Nagy Ramdan
Summary: This study utilizes machine learning models to detect depression symptoms in Arabic and English texts, and provides manually and automatically annotated tweet corpora. The study also develops an application that can detect tweets with depression symptoms and predict depression trends.
ARTIFICIAL INTELLIGENCE IN MEDICINE
(2024)