A5073968803- Particle physics theoretical and experimental studies
- Advanced Neural Network Applications
- Quantum Chromodynamics and Particle Interactions
- AI in cancer detection
- Radiomics and Machine Learning in Medical Imaging
- Domain Adaptation and Few-Shot Learning
- High-Energy Particle Collisions Research
- COVID-19 diagnosis using AI
- Multimodal Machine Learning Applications
- Gastrointestinal Bleeding Diagnosis and Treatment
- Medical Image Segmentation Techniques
- Colorectal Cancer Screening and Detection
- Functional Brain Connectivity Studies
- Brain Tumor Detection and Classification
- Medical Imaging and Analysis
- Image Retrieval and Classification Techniques
- Advanced Image and Video Retrieval Techniques
- Privacy-Preserving Technologies in Data
- Advanced Image Processing Techniques
- Topic Modeling
- Advanced Vision and Imaging
- EEG and Brain-Computer Interfaces
- Artificial Intelligence in Healthcare and Education
- Image Processing Techniques and Applications
- Advanced Neuroimaging Techniques and Applications
Chinese University of Hong Kong
2013-2025
University of Chinese Academy of Sciences
2023-2025
Institute of High Energy Physics
2004-2025
Air Force Medical University
2025
City University of Hong Kong
2002-2024
Sun Yat-sen University
2024
Chinese Academy of Sciences
2003-2024
Hunan University of Technology
2024
University of Virginia
2024
South China Agricultural University
2022-2024
Vision transformers have shown great success due to their high model capabilities. However, remarkable performance is accompanied by heavy computation costs, which makes them unsuitable for real-time applications. In this paper, we propose a family of high-speed vision named Efficient ViT. We find that the speed existing transformer models commonly bounded memory inefficient operations, especially tensor reshaping and element-wise functions in MHSA. Therefore, design new building block with...
Wireless capsule endoscopy (WCE) enables physicians to examine the digestive tract without any surgical operations, at cost of a large volume images be analyzed. In computer-aided diagnosis WCE images, main challenge arises from difficulty robust characterization images. This study aims provide discriminative description and assist recognize polyp automatically.We propose novel deep feature learning method, named stacked sparse autoencoder with image manifold constraint (SSAEIM), polyps in...
Object detection on the drone faces a great diversity of challenges such as small object inference, background clutter and wide viewpoint. In contrast to traditional problem in computer vision, bird-like angle can not be transplanted directly from common-in-use methods due special texture sky's view. However, lack comprehensive data set, number algorithms that focus using captured by drones is limited. So VisDrone team gathered massive set organized Vision Meets Drones: A Challenge...
Domain Adaptive Object Detection (DAOD) leverages a labeled domain to learn an object detector generalizing novel free of annotations. Recent advances align class-conditional distributions by narrowing down cross-domain prototypes (class centers). Though great success, they ignore the significant within-class variance and domain-mismatched semantics within training batch, leading sub-optimal adaptation. To overcome these challenges, we propose SemantIc-complete Graph MAtching (SIGMA)...
Estimating the 3D position and orientation of objects in environment with a single RGB camera is critical challenging task for low-cost urban autonomous driving mobile robots. Most existing algorithms are based on geometric constraints 2D-3D correspondence, which stems from generic 6D object pose estimation. We first identify how ground plane provides additional clues depth reasoning detection scenes. Based this observation, we then improve processing anchors introduce novel neural network...
An increasing number of public datasets have shown a marked impact on automated organ segmentation and tumor detection. However, due to the small size partially labeled problem each dataset, as well limited investigation diverse types tumors, resulting models are often segmenting specific organs/tumors ignore semantics anatomical structures, nor can they be extended novel domains. To address these issues, we propose CLIP-Driven Universal Model, which incorporates text embedding learned from...
Visual prompt engineering is a fundamental methodology in the field of visual and image artificial general intelligence. As development large vision models progresses, importance becomes increasingly evident. Designing suitable prompts for specific tasks has emerged as meaningful research direction. This review aims to summarize methods employed computer domain engineering, exploring latest advancements engineering. We present influential range on these models. It our hope that this provides...
Prompt engineering is a critical technique in the field of natural language processing that involves designing and optimizing prompts used to input information into models, aiming enhance their performance on specific tasks. With recent advancements large prompt has shown significant superiority across various domains become increasingly important healthcare domain. However, there lack comprehensive reviews specifically focusing medical field. This review will introduce latest advances for...
U-Net has become a cornerstone in various visual applications such as image segmentation and diffusion probability models. While numerous innovative designs improvements have been introduced by incorporating transformers or MLPs, the networks are still limited to linearly modeling patterns well deficient interpretability. To address these challenges, our intuition is inspired impressive results of Kolmogorov-Arnold Networks (KANs) terms accuracy interpretability, which reshape neural network...
Is there a common structural and functional cortical architecture that can be quantitatively encoded precisely reproduced across individuals populations? This question is still largely unanswered due to the vast complexity, variability, nonlinearity of cerebral cortex. Here, we hypothesize effectively represented by group-wise consistent fiber connections take novel data-driven approach explore architecture. We report dense map 358 landmarks, named Dense Individualized Common...
Wireless capsule endoscopy (WCE) enables noninvasive and painless direct visual inspection of a patient's whole digestive tract, but at the price long time reviewing large amount images by clinicians. Thus, an automatic computer-aided technique to reduce burden physicians is highly demanded. In this paper, we propose novel color feature extraction method discriminate bleeding frames from normal ones, with further localization regions. Our proposal based on twofold system. First, make full...
Ulcer is one of the most common symptoms many serious diseases in human digestive tract. Especially for ulcers small bowel where other procedures cannot adequately visualize, wireless capsule endoscopy (WCE) increasingly being used diagnosis and clinical management. Because WCE generates large amount images from whole process inspection, computer-aided detection ulcer considered an indispensable relief to clinicians. In this paper, a two-staged fully automated system proposed detect images....
Convoluted cortical folding and neuronal wiring are 2 prominent attributes of the mammalian brain. However, macroscale intrinsic relationship between these general cross-species attributes, as well underlying principles that sculpt architecture cerebral cortex, remains unclear. Here, we show axonal fibers connected to gyri significantly denser than those sulci. In human, chimpanzee, macaque brains, a dominant fraction were found be gyri. This finding has been replicated in range brains via...
Wireless capsule endoscopy (WCE) needs computerized method to reduce the review time for its large image data. In this paper, we propose an improved bag of feature (BoF) assist classification polyps in WCE images. Instead utilizing a single scale-invariant transform (SIFT) traditional BoF method, extract different textural features from neighborhoods key points and integrate them together as synthetic descriptors carry out tasks. Specifically, study influence number visual words, patch size...
Solar flares produce radiation which can have an almost immediate effect on the near-Earth environment, making it crucial to forecast in order mitigate their negative effects. The number of published approaches flare forecasting using photospheric magnetic field observations has proliferated, with varying claims about how well each works. Because different analysis techniques and data sets used, is essentially impossible compare results from literature. This problem exacerbated by low event...
Purpose Prostate cancer classification has a significant impact on the prognosis and treatment planning of patients. Currently, this is based Gleason score analysis biopsied tissues, which neither accurate nor risk free. This study aims to learn discriminative features in prostate images assist physicians classifying automatically. Methods We develop novel multiparametric magnetic resonance transfer learning (MPTL) method automatically stage cancer. first establish deep convolutional neural...
Background Accurate prediction of radiation toxicity healthy organs‐at‐risks ( OAR s) critically determines the therapy RT ) success. The existing dose–volume histogram‐based metric may grossly under/overestimate therapeutic after 27% in liver and 50% head‐and‐neck . We propose novel paradigm for by leveraging enormous potential deep learning go beyond dose/volume histograms. Experimental Design employed a database 125 stereotactic body SBRT cases with follow‐up data to train learning‐based...
Segmentation of liver in abdominal computed tomography (CT) is an important step for radiation therapy planning hepatocellular carcinoma. Practically, a fully automatic segmentation remains challenging because low soft tissue contrast between and its surrounding organs, highly deformable shape. The purpose this work to develop novel superpixel-based boundary sensitive convolutional neural network (SBBS-CNN) pipeline automated segmentation. entire CT images were first partitioned into...