A5041698837- Advanced Neural Network Applications
- Radiomics and Machine Learning in Medical Imaging
- COVID-19 diagnosis using AI
- AI in cancer detection
- Medical Image Segmentation Techniques
- Domain Adaptation and Few-Shot Learning
- Video Surveillance and Tracking Methods
- Cancer Genomics and Diagnostics
- Prenatal Screening and Diagnostics
- Face recognition and analysis
- Retinal Imaging and Analysis
- Brain Tumor Detection and Classification
- Human Pose and Action Recognition
- Epigenetics and DNA Methylation
- Myxozoan Parasites in Aquatic Species
- Aquaculture disease management and microbiota
- Digital Imaging for Blood Diseases
- Glaucoma and retinal disorders
- Muscle Physiology and Disorders
- Renal cell carcinoma treatment
- Electrospun Nanofibers in Biomedical Applications
- Neutrophil, Myeloperoxidase and Oxidative Mechanisms
- Cell Image Analysis Techniques
- Forensic and Genetic Research
- Fetal and Pediatric Neurological Disorders
Fuzhou University
2022-2024
Northwest A&F University
2023-2024
Xiamen University
2018-2022
Shangrao Normal University
2022
Chinese University of Hong Kong
2020-2021
Western University
2021
First Hospital of Shanxi Medical University
2019
Shanxi Medical University
2019
Shanxi Academy of Medical Sciences
2019
East China University of Science and Technology
2014
Retinal vessel segmentation is a key step towards the accurate visualization, diagnosis, early treatment and surgery planning of ocular diseases. Recently, deep learning based retinal methods have reached state-of-the-art performance. Due to extreme variations in morphology vessels against noisy background, these still issues dealing with small thin vessels, low discriminative ability at optic disk area, etc. In this paper, we proposed U-Net-like model weighted attention mechanism skip...
Retinal vessel segmentation is a critical procedure towards the accurate visualization, diagnosis, early treatment, and surgery planning of ocular diseases. Recent deep learning-based approaches have achieved impressive performance in retinal segmentation. However, they usually apply global image pre-processing take whole images as input during network training, which two drawbacks for First, these methods lack utilization local patch information. Second, overlook geometric constraint that...
Granulomatous diseases caused by Nocardia seriously endanger the health of cultured fish. These bacteria are widely distributed, but prevention and treatment methods very limited. Chronic granulomatous inflammation is an important pathological feature infection. However, molecular mechanisms granuloma formation chronic still unclear. Constructing a infection model key to exploring pathogenesis disease. In this study, we established in liver largemouth bass (Micropterus salmoides) assessed...
Due to the effective multi-scale feature fusion capabilities of Path Aggregation FPN (PAFPN), it has become a widely adopted component in YOLO-based detectors. However, PAFPN struggles integrate high-level semantic cues with low-level spatial details, limiting its performance real-world applications, especially significant scale variations. In this paper, we propose MHAF-YOLO, novel detection framework featuring versatile neck design called Multi-Branch Auxiliary (MAFPN), which consists two...
Person re-identification (re-ID) is a challenging instance retrieval problem, especially when identity annotations are not available for training. Although modern deep re-ID approaches have achieved great improvement, it still difficult to optimize the model and learn discriminative person representation without in training data. To address this challenge, study considers problem of unsupervised introduces novel approach solve by leveraging virtual real Our includes two components:...
Summary According to statistics, kidney cancer is one of the most deadly cancer. An early and accurate diagnosis can significantly increase cure rate. Accurate segmentation tumors in CT images plays an important role diagnosis. However, it a challenging task due many different aspects, such as low contrast, irregular motion, diverse shapes, sizes. For solving this issue, we proposed SE‐R esNeXT U ‐Net (SERU) model study, which takes advantages SE‐Net, ResNeXT U‐Net. Besides, implement our...
We developed genetic-epigenetic tissue mapping (GETMap) to determine the composition of plasma DNA carrying genetic variants not present in constitutional genome through comparing their methylation profiles with relevant tissues. validated this approach by showing that, pregnant women, circulating fetal-specific alleles was entirely placenta-derived. In lung transplant recipients, we showed at 72 hr after transplantation, contributed only a median 17% donor-specific alleles, and...
Mesenchymal stem cell (MSC) transplantation by intramyocardial injection has been proposed as a promising therapy strategy for cardiac repair after myocardium infarction. However, low retention and survival of grafted MSCs hinder its further application. In this study, copolymer with N-isopropylacrylamide/acrylic acid/2-hydroxylethyl methacrylate-poly(ɛ-caprolactone) ratio 88:9.6:2.4 was bioconjugated type I collagen to construct novel injectable thermosensitive hydrogel. The biocompatible...
Summary Multi‐organ segmentation is a critical prerequisite for many clinical applications. Deep learning‐based approaches have recently achieved promising results on this task. However, they heavily rely massive data with multi‐organ annotated, which labor‐ and expert‐intensive thus difficult to obtain. In contrast, single‐organ datasets are easier acquire, well‐annotated ones publicly available. It leads the partially labeled issue: How learn unified model from several datasets?...
Anchor-based deep methods are the most widely used for face detection and have reached state-of-the-art result. Compared with anchor-based that estimates bounding-box rely on some pre-defined anchor boxes, anchor-free perform localization by predicting offsets of a pixel inside to its outside boundaries whose accuracies much more precise. However, suffer drawback low recall-rate mainly because 1) only using single scale features lead miss small faces, 2) highly intra-class imbalance problem...
Few-shot learning can potentially learn the target knowledge in extremely few data regimes. Existing few-shot medical image segmentation methods fail to consider global anatomy correlation between support and query sets. They generally adopt a weak one-way information transmission that not fully explore segment data. To address this problem, we propose novel Symmetrical Supervision network based on traditional two-branch methods. We raise two main contributions: (1) The Mechanism is...
Abstract Facioscapulohumeral muscular dystrophy type 1 (FSHD1) patients exhibit marked variability in both age at onset (AAO) and disease severity. Early FSHD1 are an increased risk of severe weakness, early has been tentatively linked to the length D4Z4 repeat units (RUs) methylation levels. The present study explored potential relationships among genetic characteristics, AAO severity FSHD1. This retrospective observational cohort was conducted Fujian Neuromedical Centre (FNMC) China....