A5101411299- Advanced Neural Network Applications
- Radiomics and Machine Learning in Medical Imaging
- AI in cancer detection
- Medical Image Segmentation Techniques
- Brain Tumor Detection and Classification
- Medical Imaging and Analysis
- Face and Expression Recognition
- COVID-19 diagnosis using AI
- Face recognition and analysis
- Energy Efficient Wireless Sensor Networks
- Multimodal Machine Learning Applications
- Renal cell carcinoma treatment
- Cell Image Analysis Techniques
- Anomaly Detection Techniques and Applications
- Ferroptosis and cancer prognosis
- Industrial Vision Systems and Defect Detection
- Human Pose and Action Recognition
- Biometric Identification and Security
- Indoor and Outdoor Localization Technologies
- Advanced Image and Video Retrieval Techniques
- Network Security and Intrusion Detection
- RNA modifications and cancer
- Image Retrieval and Classification Techniques
- Cancer Genomics and Diagnostics
- Cutaneous Melanoma Detection and Management
Soochow University
2020-2025
China University of Geosciences
2025
Shenzhen Institutes of Advanced Technology
2024
Lenovo (China)
2017-2024
Jinan University
2024
Guangzhou University of Chinese Medicine
2024
Jiangmen Wuyi Traditional Chinese Medicine Hospital
2024
University of Science and Technology Beijing
2024
Zhejiang University
2023
Tongji University
2023
The capabilities of (I) learning transferable knowledge across domains; and (II) fine-tuning the pre-learned base towards tasks with considerably smaller data scale are extremely important. Many existing transfer techniques supervised approaches, among which deep has demonstrated power domain transferrable large network trained on massive amounts labeled data. However, in many biomedical tasks, both corresponding label can be very limited, where unsupervised capability is urgently needed. In...
Stocks that are fundamentally connected with each other tend to move together. Considering such common trends is believed benefit stock movement forecasting tasks. However, signals not trivial model because the connections among stocks physically presented and need be estimated from volatile data. Motivated by this observation, we propose a framework incorporates inter-connection of firms forecast prices. To effectively utilize large set fundamental features, further design novel pipeline....
Hepatic ischemia-reperfusion (IR) injury is a serious clinical problem that complicates liver resection and transplantation. Despite recent advances in understanding of the pathophysiology hepatic IR injury, effective interventions therapeutics are still lacking. Here, we examined role transient receptor potential melastatin 2 (TRPM2), Ca2+-permeable, non-selective cation channel, mediating injury. Our data showed TRPM2 deficiency attenuated IR-induced dysfunction, inflammation, cell death...
Delineating the extent of urban villages (UVs) is crucial for effective planning and management, as well providing targeted policy financial support. Unlike field surveys, interpretation satellite imagery provides an efficient, near real-time, objective means mapping UV. However, current research efforts predominantly concentrate on individual cities, resulting in a scarcity interpretable UV maps numerous other cities. This gap availability not only hinders public awareness distribution...
Cuproptosis was characterized as a novel type of programmed cell death. Recently, however, the role cuproptosis-related long noncoding RNAs (CRLs) in tumors has not yet been studied. Identifying predictive CRL signature hepatocellular carcinoma (HCC) and investigating its putative molecular function were goals this work. Initially, Pearson’s test used to assess relationship between lncRNAs cuproptosis-associated genes obtained from HCC data The Cancer Genome Atlas (TCGA). By implementing...
With the advances in single-cell sequencing techniques, numerous analytical methods have been developed for delineating cell development. However, most are based on Euclidean space, which would distort complex hierarchical structure of differentiation. Recently, acting hyperbolic space proposed to visualize structures RNA-seq (scRNA-seq) data and proven be superior space. these fundamental limitations not optimized highly sparse count data. To address limitations, we propose scDHMap, a...
Three hundred and twelve acute large vessel occlusion stroke patients who underwent mechanical thrombectomy at Changshu Hospital of Traditional Chinese Medicine, Suzhou Ninth People's Hospital, the Second Affiliated Soochow University from July 2016 to December 2023 were retrospectively collected. The divided into effective recanalization group(mRS<3) ineffective (mRS≥3) group based on 90-day modified Rankin scale(mRS) scores. Serum magnesium levels categorized tertiles: T1(Mg2+≤0.75...
In this paper, we propose a novel learned visual code-book (LVC) for 3D face recognition. our method, first extract intrinsic discriminative information embedded in faces using Gabor filters, then K-means clustering is adopted to learn the centers from filter response vectors. We construct LVC by these centers. Finally represent based on and achieve recognition nearest neighbor (NN) classifier. The novelty of paper comes 1) apply textons methods into recognition; 2) encompass efficiency...
Objective: In order to understand the epidemic trend of COVID-19 and evaluate effect prevention control, this study aims dynamics in Chinese mainland from January 16 February 14, 2020. Methods: The daily number new confirmed cases detected by nucleic acid was collected National Health Commission 16, 2020 analysis included curve cases, multiple for period-over-period, fixed-base, period-over-period growth rate cases. Results: From 2020, cumulative 50 031, including 37 930 Hubei Province, 22...
Deep learning has the potential to improve diagnostic accuracy and efficiency in medical image recognition. In current study, we developed a deep algorithm assessed its performance discriminating melanoma from nevus using whole-slide pathological images (WSIs). The was trained validated set of 781 WSIs (86 melanomas, 695 nevi) PLA General Hospital. tested on an independent test 104 (29 75 Tianjin Chang Zheng same also diagnostically classified by 7 expert dermatopathologists. receiver...
Subway vehicle roofs must be inspected when entering and exiting the depot to ensure safe subway operations. This paper presents an improved method for detecting foreign objects on based YOLOv7 algorithm. First, we capture images of using a line-scan camera at entrance exit, creating dataset roof objects. Subsequently, address shortcomings algorithm by introducing Ghost module, weighted bidirectional feature pyramid network (WBiFPN), Wise intersection over union (WIoU) bounding-box...
In this paper, we present a novel iris recognition method based on learned ordinal features.Firstly, taking full advantages of the properties textures, new representation regional measure encoding is presented, which provides an over-complete feature set for learning. Secondly, Similarity Oriented Boosting (SOBoost) algorithm proposed to train efficient and stable classifier with small features. Compared Adaboost, SOBoost advantageous in that it operates similarity oriented training samples,...
Nowadays, in machine learning based intrusion detection systems, ensemble is a commonly adopted method to improve the accuracy. Unfortunately, existing works have not considered accumulation and reuse of historical knowledge, as well sensitivity model different types attacks, which leads low To address issue, this article proposes on sustainable learning. In training stage, by taking individual classifiers probability output classification confidence data, we build multi-class regression...
Automated segmentation of kidney and tumor from 3D CT scans is necessary for the diagnosis, monitoring, treatment planning disease.In this paper, we describe a two-stage framework based on fully convolutional network (FCN).The first stage preliminarily locate cut off irrelevant background to reduce class imbalance computation cost.Then second precisely segment cropped patch.The proposed method achieves 98.05% 83.70% Dice score validation set MICCAI 2019 KiTS Challenge.
Abstract In addition to presenting significant diagnostic and treatment challenges, lung adenocarcinoma (LUAD) is the most common form of cancer. Using scRNA-Seq bulk RNA-Seq data, we identify three genes referred as HMR, FAM83A, KRT6A these are related necroptotic anoikis-related gene expression. Initial validation, conducted on GSE50081 dataset, demonstrated model's ability categorize LUAD patients into high-risk low-risk groups with survival differences. This model was further applied...
Automatic liver segmentation from abdominal Computed Tomography (CT) is an important step for hepatic disease diagnosis. It a challenging task owing to the similarity between and its adjacent organs low contrast of texture (e.g. tumors blood veins). In this paper, we propose cascaded structure automatically segment in CT scans. First, train fully convolutional neural network (FCN) coarse segmentation; second, make comparative study performance different classical models as post-processing...