Article
Computer Science, Artificial Intelligence
Chunfeng Lian, Mingxia Liu, Li Wang, Dinggang Shen
Summary: The proposed multi-task weakly-supervised attention network (MWAN) can jointly regress multiple clinical scores through automatically identifying subject-specific discriminative brain locations, achieving superior performance in predicting dementia progression on public AD datasets.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Waseem Ullah, Tanveer Hussain, Fath U. Min Ullah, Khan Muhammad, Mahmoud Hassaballah, Joel J. P. C. Rodrigues, Sung Wook Baik, Victor Hugo C. de Albuquerque
Summary: The main challenge faced by video-based real-world anomaly detection systems is accurately learning irregular, complicated, diverse, and heterogeneous unusual events. To address this, a weakly supervised graph neural-network-assisted video anomaly detection framework called AD-Graph is proposed. It extracts 3D visual and motion features and represents them in a language-based knowledge graph format to identify temporal information. It applies a robust clustering strategy to group meaningful neighborhoods of the graph and uses spectral filters and graph theory to detect anomalous events. Extensive experimental results show improvements over state-of-the-art models on challenging datasets.
INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Yongfei Zhang, Ling Dong, Hong Yang, Linbo Qing, Xiaohai He, Honggang Chen
Summary: The deep learning-based image super-resolution method proposes a new blind image super-resolution approach to simulate different degradation methods, thus adapting better to various scenarios.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Multidisciplinary Sciences
Soo Hyun Cho, Sookyoung Woo, Changsoo Kim, Hee Jin Kim, Hyemin Jang, Byeong C. Kim, Si Eun Kim, Seung Joo Kim, Jun Pyo Kim, Young Hee Jung, Samuel Lockhart, Rik Ossenkoppele, Susan Landau, Duk L. Na, Michael Weiner, Seonwoo Kim, Sang Won Seo
Summary: This study aimed to construct a disease course model from preclinical AD to AD dementia, finding that ADAS-cog 13 scores decreased most rapidly in women APOE epsilon 4 carriers and most slowly in men APOE epsilon 4 non-carriers. The results suggest that both sex and APOE epsilon 4 status have an impact on the progression of AD.
SCIENTIFIC REPORTS
(2021)
Article
Computer Science, Artificial Intelligence
Bongyeong Koo, Han-Soo Choi, Myungjoo Kang
Summary: In this study, a novel weakly supervised object localization (WSOL) method is proposed, which combines a spatial attention branch and a refinement branch for accurate localization. By enhancing spatial information and considering spatial relationships, the proposed method achieves state-of-the-art performance on the CUB-200-2011 and ILSVRC 2012 datasets. The method also demonstrates efficiency with lightweight trainable parameters.
IMAGE AND VISION COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Jie Lin, Yu Zhan, Wan-Lei Zhao
Summary: This paper proposes a compact instance level feature representation using two CNN pipelines to localize potential instances and generate distinctive features, considering factors such as sensitivity to unknown categories, distinctiveness to different instances, and capability of localizing an instance in an image. This method is suitable for large-scale image collections and is the first work to build instance level representation based on weakly supervised object detection.
Article
Computer Science, Artificial Intelligence
Wanchun Sun, Xin Feng, Hui Ma, Jingyao Liu
Summary: This paper proposes a method that learns the semantic relationships between different blocks using a graph structure to address the issue of determining semantic association in transformer-based weakly supervised semantic segmentation. Experimental results on the PASCAL VOC2012 dataset demonstrate the significant performance improvement of the proposed method in the WSSS task, outperforming other state-of-the-art transformer-based methods.
COMPLEX & INTELLIGENT SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Byeongjoon Kim, Hyunjung Shim, Jongduk Baek
Summary: In this study, a weakly-supervised denoising framework is proposed to generate paired original and noisier CT images from unpaired CT images using a physics-based noise model. The experimental results demonstrate that the method achieves remarkable performances in diagnostic image quality, even superior to fully-supervised CT denoising in terms of signal detectability.
MEDICAL IMAGE ANALYSIS
(2021)
Article
Computer Science, Artificial Intelligence
Junsuk Choe, Seungho Lee, Hyunjung Shim
Summary: The proposed attention-based dropout layer efficiently locates the entire object, improving weakly supervised single object localization and semantic segmentation accuracy. This method utilizes two key components: hiding the most discriminative part and highlighting the informative region.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2021)
Article
Computer Science, Information Systems
Yi Zhong, Jia-Hui Pan, Haoxin Li, Wei-Shi Zheng
Summary: This paper proposes a weakly supervised method that uses global motion and local fine-grained features from current action videos to predict the next action label without the need for specific scene context labels.
FRONTIERS OF COMPUTER SCIENCE
(2023)
Article
Computer Science, Hardware & Architecture
Yu-e Lin, Houguo Li, Xingzhu Liang, Mengfan Li, Huilin Liu
Summary: This paper proposes a novel method for weakly supervised semantic segmentation, which utilizes Attention Activation Remodulation (AAR) scheme and Feature Pixel Extraction Module (FPEM). The AAR scheme re-arranges the distribution of important features from channel and space perspectives, improving the activation response for segmentation. Experiments demonstrate the effectiveness of this method.
JOURNAL OF SUPERCOMPUTING
(2023)
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
Songxiang Yang, Lin Ma, Xuezhi Tan
Summary: This paper proposes an end-to-end weakly supervised class-agnostic image retrieval method based on convolutional neural networks. The method preprocesses and clusters the database images to avoid mismatches caused by background mixing, and shows better performance on multiple datasets.
IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTING
(2022)
Article
Computer Science, Artificial Intelligence
Zi-Wei Li, Shi-Bin Xuan, Xue-Dong He, Li Wang
Summary: Medical image segmentation is important in computer-aided diagnosis and intelligent medical treatment. Class activation map (CAM) is a weakly supervised segmentation technology that can achieve image segmentation without pixel-level label training. However, CAM's performance is affected by global average pooling (GAP) as it fails to accurately demarcate the boundaries of target regions. To address this issue, a global weighted average pooling network that fuses grayscale information is proposed.
IET IMAGE PROCESSING
(2023)
Article
Computer Science, Information Systems
Gautam Kumar, Prateek Keserwani, Partha Pratim Roy, Debi Prosad Dogra
Summary: The paper introduces a logo detection method utilizing weakly supervised learning with CNN to generate a deep saliency map for logo region detection. The method fine-tunes AlexNet, CaffeNet, and VGGNet deep architectures for classification and achieves a mean average precision of 75.83% on the FlickrLogos-32 dataset, outperforming existing fully supervised methods.
MULTIMEDIA TOOLS AND APPLICATIONS
(2021)
Article
Geosciences, Multidisciplinary
Youcun Liu, Haohong Huang, Lihong Meng, Mingxia Liu, Zidan Wu, Tao Liu, David Labat
Summary: Based on the analysis of the MODIS NDVI and DEM data for the upper reaches of the Ganjiang River Basin in China from 2000 to 2020, this study examined the temporal and spatial evolution of vegetation coverage and its relationship with terrain factors. The results showed that the highest vegetation coverage was found at an elevation of 750-1,000 meters, with an average of 83.54%. Vegetation coverage also increased with higher elevation and slope. Human activities mainly affected areas at low elevation and small slopes where cities and towns are located.
FRONTIERS IN EARTH SCIENCE
(2023)
Article
Health Care Sciences & Services
Chengcheng Wang, Limei Zhang, Jinshan Zhang, Lishan Qiao, Mingxia Liu
Summary: In this study, a multiview functional brain network fusion strategy through joint embedding was proposed for fMRI-based ASD identification. The experimental results showed that the proposed method outperformed other methods in automated ASD diagnosis and discovered potential biomarkers for ASD diagnosis.
JOURNAL OF PERSONALIZED MEDICINE
(2023)
Article
Neurosciences
Yuqi Fang, Guy G. Potter, Di Wu, Hongtu Zhu, Mingxia Liu
Summary: In this article, a dual-expert fMRI harmonization (DFH) framework is proposed for automated MDD diagnosis. The framework can utilize data from a labeled source domain/site and two unlabeled target domains to reduce data distribution differences. Experimental results demonstrate the superiority of our method in identifying MDD and suggest potential biomarkers for fMRI-related MDD diagnosis.
HUMAN BRAIN MAPPING
(2023)
Article
Computer Science, Interdisciplinary Applications
Fan Wang, Xuanang Xu, Defu Yang, Ronald C. Chen, Trevor J. Royce, Andrew Wang, Jun Lian, Chunfeng Lian
Summary: This paper proposes an asymmetric multi-task network integrating dynamic cross-task representation adaptation for accurate and efficient co-segmentation of prostate bed and organs at risk from CT images. The method adapts the hierarchical feature representations from the source task to match up with the more challenging target task. On a real-patient dataset, the method achieves state-of-the-art results of co-segmentation.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Neurosciences
Mingxia Liu, Mo Li, Jing He, Yi He, Jian Yang, Zuoli Sun
Summary: This study investigated the levels of chiral amino acids in the peripheral serum of Alzheimer's disease (AD) patients and found that D-proline, D-aspartate, and D-phenylalanine were altered in AD patients, which may serve as novel biomarkers for AD. Additionally, the level of D-aspartate was associated with the severity of AD.
JOURNAL OF ALZHEIMERS DISEASE
(2023)
Article
Allergy
Bilal Haider Shamsi, Haiyuan Chen, Xiong Yang, Mingxia Liu, Yonglin Liu
Summary: This study found that SNPs in GSDMB gene are associated with susceptibility to allergic rhinitis (AR), with rs4795400 showing a protective effect in overall, males, people with BMI ≤ 24, and those living in wind-blown sand area. rs2305479 is associated with reduced risk of AR in males, while rs12450091 is a risk factor for AR in people living in the loess hilly region. The findings suggest that GSDMB polymorphisms are correlated with AR susceptibility.
Article
Public, Environmental & Occupational Health
Tingting Wang, Qi Gao, Yuanyuan Yao, Ge Luo, Tao Lv, Guangxin Xu, Mingxia Liu, Jingpin Xu, Xuejie Li, Dawei Sun, Zhenzhen Cheng, Ying Wang, Chaomin Wu, Ruiyu Wang, Jingcheng Zou, Min Yan
Summary: Using Mendelian randomization, this study found that obesity may cause iron deficiency anemia.
FRONTIERS IN PUBLIC HEALTH
(2023)
Article
Engineering, Aerospace
Jintao Chen, Mingxia Liu, Yuxiang Zhu, Kairu Jin, Zhenyu Tian, Lijun Yang, Chong -Wen Zhou
Summary: The chemical kinetics of hydrogen atom (H-atom) abstraction reactions from nor-bornadiene (NBD) by five radicals (H, O(3P), OH, CH3, and HO2), and the unimolecular reac-tions of three NBD derived radicals, were studied through high-level ab-initio calculations. The results show that the H-atom abstraction reactions from the a-carbon atom of NBD are the most critical channels at low temperatures. Total rate constants for H-atom abstraction reactions by OH radical are also the fastest among all of the reaction channels investigated at the temperature range from 298.15 to 2000 K.
PROPULSION AND POWER RESEARCH
(2023)
Article
Computer Science, Interdisciplinary Applications
Caiwen Jiang, Yongsheng Pan, Zhiming Cui, Dong Nie, Dinggang Shen
Summary: PET is a widely used nuclear medical imaging technique for tumor detection and brain disease diagnosis. To reduce the radiation risk while maintaining the imaging quality, we proposed a novel approach to estimate high-quality PET images from low-dose PET images. Our method utilizes both paired and unpaired images in a semi-supervised framework and incorporates region-adaptive normalization and a structural consistency constraint to address task-specific challenges. The experiments on real human PET images show that our approach achieves state-of-the-art performance both quantitatively and qualitatively.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Chemistry, Physical
Chen Wang, Chaofan Zeng, Haiyue Ning, Fengnan Li, Mingxia Liu, Kewei Xu, Fei Ma
Summary: Pulsed laser deposition and two-step thermal annealing are used to fabricate passivation layers on a-IGZO TFTs. The influences of the passivation layers on electrical properties are investigated. The combination of HC-PLD and oxygen plasma is employed to solve the oxygen deficiency in certain thin films. Furthermore, two-step annealing enhances the carrier mobility and improves the on-off performance.
JOURNAL OF ALLOYS AND COMPOUNDS
(2023)
Article
Anesthesiology
Yue Ming, Fengjiang Zhang, Yuanyuan Yao, Zhenzhen Cheng, Lina Yu, Dawei Sun, Kai Sun, Yang Yu, Mingxia Liu, Longfei Ma, Yuxin HuangYang, Min Yan
Summary: The study aimed to investigate whether large volume acute normovolemic hemodilution (L-ANH) can reduce perioperative allogeneic blood transfusion in patients undergoing cardiac surgery with cardiopulmonary bypass (CPB) compared to moderate acute normovolemic hemodilution (M-ANH). The results showed that L-ANH was associated with a lower incidence of perioperative red blood cell transfusion and postoperative excessive bleeding compared to M-ANH.
JOURNAL OF CLINICAL ANESTHESIA
(2023)
Article
Computer Science, Artificial Intelligence
Yongheng Sun, Duwei Dai, Qianni Zhang, Yaqi Wang, Songhua Xu, Chunfeng Lian
Summary: This paper presents a skin lesion segmentation model called MSCA-Net, which can utilize multi-scale contextual information in images. Comprehensive experimental results on the public datasets ISIC 2017, ISIC 2018, and PH2 demonstrate that our proposed method outperforms other state-of-the-art methods, showing its effectiveness in skin lesion segmentation.
PATTERN RECOGNITION
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Bao Wang, Yongsheng Pan, Shangchen Xu, Yi Zhang, Yang Ming, Ligang Chen, Xuejun Liu, Chengwei Wang, Yingchao Liu, Yong Xia
Summary: Image-to-image translation techniques can accurately generate synthetic cerebral blood volume (CBV) maps from standard MRI scans, which can improve the clinical evaluation of brain tumors.
Article
Computer Science, Information Systems
Ziyang Chen, Yongsheng Pan, Yong Xia
Summary: Glaucoma affects irreversible blindness, and segmenting the optic disc (OD) and optic cup (OC) on fundus images is key in screening for this disease. However, training a segmentation model that can be deployed across different healthcare centers remains challenging due to variations in image tone, contrast, and brightness. To address this, a novel unsupervised domain adaptation method called RDR-Net is proposed, which includes three modules designed to alleviate the domain gap. Evaluation against other models on four fundus image datasets demonstrates that RDR-Net excels in both performance and generalization ability.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2023)
Article
Engineering, Electrical & Electronic
Jiazhen Wang, Yan Yang, Heran Yang, Chunfeng Lian, Zongben Xu, Jian Sun
Summary: In this paper, we propose a model-driven graph transformer (MD-GraphFormer) for fast multi-contrast MR imaging, which incorporates the physical constraints of MRI and investigates the complementary information among multi-contrast MR images using graph structure and attention mechanism. Extensive experiments demonstrate that the proposed MD-GraphFormer outperforms the previous multi-contrast MRI reconstruction methods in multi-coil imaging settings.
IEEE TRANSACTIONS ON COMPUTATIONAL IMAGING
(2023)