A5027264311- Retinal Imaging and Analysis
- Retinal and Optic Conditions
- Image Retrieval and Classification Techniques
- Glaucoma and retinal disorders
- Digital Imaging for Blood Diseases
- Imbalanced Data Classification Techniques
- Click Chemistry and Applications
- Retinal Diseases and Treatments
- Catalytic C–H Functionalization Methods
- Image Processing Techniques and Applications
- Cyclopropane Reaction Mechanisms
- Ferroelectric and Piezoelectric Materials
- Ocular Diseases and Behçet’s Syndrome
- Radical Photochemical Reactions
- Dielectric materials and actuators
- Chemistry and Chemical Engineering
- Advanced MIMO Systems Optimization
- Musicians’ Health and Performance
- Advanced Image Processing Techniques
- Carbon dioxide utilization in catalysis
- Wireless Communication Networks Research
- Advanced Vision and Imaging
- CO2 Reduction Techniques and Catalysts
- AI in cancer detection
- Advanced Sensor and Control Systems
Chinese University of Hong Kong
2023-2025
Shantou University
2023-2025
Shantou University Medical College
2024-2025
Changchun University of Science and Technology
2023
University of Chinese Academy of Sciences
2011-2021
GP Batteries (China)
2020
Robert Bosch (Germany)
2020
Laboratoire de Chimie Moléculaire et Thioorganique
2020
Lithium Power (United States)
2020
Weatherford College
2019
Failure to recognize samples from the classes unseen during training is a major limitation of artificial intelligence in real-world implementation for recognition and classification retinal anomalies. We establish an uncertainty-inspired open set (UIOS) model, which trained with fundus images 9 conditions. Besides assessing probability each category, UIOS also calculates uncertainty score express its confidence. Our model thresholding strategy achieves F1 99.55%, 97.01% 91.91% internal...
Label noise is a common and important issue that would affect the model's performance in artificial intelligence. This study assessed effectiveness potential risks of automated label cleaning using an open-source framework, Cleanlab, multi-category datasets fundus photography optical coherence tomography, with intentionally introduced ranging from 0 to 70%. After six cycles automatic cleaning, significant improvements are achieved accuracies (3.4–62.9%) dataset quality scores (DQS,...
Using conventional resource allocation algorithms in OFDM systems, each user can employ different Modulation and Coding Scheme (MCS) on allocated subcarriers to achieve good throughput. However, the downlink transmission of LTE minimum unit for one is Scheduling Block (SB) all SB assigned must adopt same MCS. Therefore, application results degraded performance since MCS be chosen according worst SB. To solve this problem, a QoS guaranteed block algorithm proposed which takes into account...
Optical Coherence Tomography (OCT) images can provide non-invasive visualization of fundus lesions; however, scanners from different OCT manufacturers largely vary each other, which often leads to model deterioration unseen due domain shift.
An actuator with a high piezoelectric response, the ferroelectric polymer multilayer actuator, is described. The multilayers consisting of alternative poly(vinylidene fluoride-trifluoroethylene) (P(VDF-TrFE)) copolymer and relaxor fluoride-trifluoroethylene-chlorofloroethylene) (P(VDF-TrFE-CFE)) terpolymer different periodicities fixed total thickness are prepared by Langmuir-Blodgett technique. Both X-ray diffraction Raman spectroscopic measurements indicate that structure thin alternating...
Multi-modal ophthalmic image classification plays a key role in diagnosing eye diseases, as it integrates information from different sources to complement their respective performances. However, recent improvements have mainly focused on accuracy, often neglecting the importance of confidence and robustness predictions for diverse modalities. In this study, we propose novel multi-modality evidential fusion pipeline disease screening. It provides measure each modality elegantly using...
The current retinal artificial intelligence models were trained using data with a limited category of diseases and knowledge. In this paper, we present vision-language foundation model (RetiZero) knowledge over 400 fundus diseases. Specifically, collected 341,896 images paired text descriptions from 29 publicly available datasets, 180 ophthalmic books, online resources, encompassing across multiple countries ethnicities. RetiZero achieved outstanding performance various downstream tasks,...
Deep convolutional neural networks (CNNs) have been widely used for fundus image classification and achieved very impressive performance. However, the explainability of CNNs is poor because their black-box nature, which limits application in clinical practice. In this paper, we propose a novel method to search discriminative regions increase confidence features specific category, thereby helping users understand an are important CNN make particular prediction. proposed method, set...
The semantic consistency, loss of detail features, and other problems are caused by the traditional text generation image method's focus on extracting information transmitted previous layer while ignoring layer's features that lost during subsequent propagation process. Text-to-images combining blending structure attention (CAGAN) with fusion capabilities is suggested in order to address aforementioned issues. word sentence levels text's characteristics encoded using pre-trained BiLSTM...