A5101812640- Advanced Neural Network Applications
- Radiomics and Machine Learning in Medical Imaging
- AI in cancer detection
- Medical Image Segmentation Techniques
- Digital Radiography and Breast Imaging
- Caching and Content Delivery
- Remote Sensing and LiDAR Applications
- Robotics and Sensor-Based Localization
- Advanced Image and Video Retrieval Techniques
- 3D Shape Modeling and Analysis
- Polynomial and algebraic computation
- Domain Adaptation and Few-Shot Learning
- Advanced Photocatalysis Techniques
- Ammonia Synthesis and Nitrogen Reduction
- Computer Graphics and Visualization Techniques
- Medical Imaging and Analysis
- Cell Image Analysis Techniques
- Multimodal Machine Learning Applications
- Image Processing Techniques and Applications
- Computational Geometry and Mesh Generation
- Cloud Computing and Remote Desktop Technologies
- Image Enhancement Techniques
- Neurological disorders and treatments
- Parallel Computing and Optimization Techniques
- Advanced Optimization Algorithms Research
City University of Hong Kong
2021-2025
Chang'an University
2024
National University of Singapore
2022-2023
University of Science and Technology of China
2023
Southeast University
2023
Hunan University
2021-2022
National University Health System
2022
Ocean University of China
2021
Guizhou Electric Power Design and Research Institute
2019
Chinese Academy of Sciences
2017
The precise segmentation of medical images is one the key challenges in pathology research and clinical practice. However, many image tasks have problems such as large differences between different types lesions similar shapes well colors surrounding tissues, which seriously affects improvement accuracy. In this article, a novel method called Swin Pyramid Aggregation network (SwinPA-Net) proposed by combining two designed modules with Transformer to learn more powerful robust features....
Differentiable ARchiTecture Search, i.e., DARTS, has drawn great attention in neural architecture search. It tries to find the optimal a shallow search network and then measures its performance deep evaluation network. The independent optimization of networks, however, leaves room for potential improvement by allowing interaction between two networks. To address problematic issue, we propose new joint objectives novel Cyclic Search framework, dubbed CDARTS. Considering structure difference,...
Digital reconstruction of neuronal structures from 3D microscopy images is critical for the quantitative investigation brain circuits and functions. It a challenging task that would greatly benefit automatic neuron methods. In this paper, we propose novel method called SPE-DNR combines spherical-patches extraction (SPE) deep-learning (DNR). Based on 2D Convolutional Neural Networks (CNNs) intensity distribution features extracted by SPE, it determines tracing directions classifies voxels...
Accurate medical image segmentation is critical to effective treatment strategies. Existing transformer-based methods for mostly split the input into a fixed and regular grid regard cells in as vision tokens. However, not all tokens are of equal importance tasks, e.g., tumor areas must be processed higher resolution than background which can easily predicted with fewer transformer layers. In this paper, we propose simple yet efficient framework called Top-Down Transformer (TDFormer),...
Medical image segmentation is almost the most important pre-processing procedure in computer-aided diagnosis but also a very challenging task due to complex shapes of segments and various artifacts caused by medical imaging, (i.e., low-contrast tissues, non-homogenous textures). In this paper, we propose simple yet effective framework that incorporates geometric prior contrastive similarity into weakly-supervised loss-based fashion. The proposed built on point cloud provides meticulous...
In this paper, we study autonomous landing scene recognition with knowledge transfer for drones. Considering the difficulties in aerial remote sensing, especially that some scenes are extremely similar, or same has different representations altitudes, employ a deep convolutional neural network (CNN) based on and fine-tuning to solve problem. Then, LandingScenes-7 dataset is established divided into seven classes. Moreover, there still novelty detection problem classifier, address by...
Existing image captioning methods are under the assumption that training and testing data from same domain or target (i.e., lie in) accessible. However, this is invalid in real-world applications where inaccessible. In paper, we introduce a new setting called Domain Generalization for Image Captioning (DGIC), unseen learning process. We first construct benchmark dataset DGIC, which helps us to investigate models' generalization (DG) ability on domains. With support of benchmark, further...
Background: Although several studies have been conducted on the use of Artificial Intelligence (AI) in mammography (MG), research focusing diagnosis metachronous bilateral breast cancer (BC), which is typically more challenging to diagnose, remains limited. This study assessed whether AI could enhance detection BC, achieving earlier or accurate diagnoses than radiologists cases contralateral BC. Methods: The included patients who underwent unilateral BC surgery and subsequently developed...
Purpose: We aimed to develop a novel interpretable artificial intelligence (AI) model algorithm focusing on automatic detection and classification of various patterns calcification distribution in mammographic images using unique graph convolution approach. Materials methods: Images from 292 patients, which showed calcifications according the reports diagnosed breast cancers, were collected. The distributions classified as diffuse, segmental, regional, grouped, or linear. Excluded mammograms...
Medical image registration is a critical task that estimates the spatial correspondence between pairs of images. However, current traditional and deep-learning-based methods rely on similarity measures to generate deforming field, which often results in disproportionate volume changes dissimilar regions, especially tumor regions. These can significantly alter size underlying anatomy, limits practical use clinical diagnosis. To address this issue, we have formulated with tumors as constraint...
In-memory key-value store is a crucial building block of large-scale web architecture. Given the growth data volume and need for low-latency responses, cost-effective storage expansion fast large-message processing are major challenges. In this paper, we explore design middleware that takes advantage modern NVMe SSDs RDMA interconnects to achieve high performance without excessive DRAM deployment. We propose an all-in-userland approach improve plane efficiency. Both interfaced directly from...
The morphology and distribution of microcalcifications in a cluster are the most important characteristics for radiologists to diagnose breast cancer. However, it is time-consuming difficult identify these characteristics, there also lacks effective solutions automatic characterization. In this study, we proposed multi-task deep graph convolutional network (GCN) method characterization mammograms. Our transforms into node classification problem learns representations concurrently. Through...
Tree barrier analysis based on laser point cloud has been widely used in power grid industry. This grammar uses clustering idea and DBSCAN method to propose a line segmentation algorithm airborne LiDAR. By using this algorithm, we can realized the of single line. And it is suitable for absence small range lines. Finally, carry out tree by segmented feasibility verified comparing with manual detection.
In this paper, we study scene image recognition with knowledge transfer for drone navigation. We divide navigation scenes into three macro-classes, namely outdoor special (OSSs), the space from indoors to outdoors or transitional (TSs), and others. However, there are difficulties in how recognize TSs, end, employ deep convolutional neural network (CNN) based on transfer, techniques augmentation, fine tuning solve issue. Moreover, is still a novelty detection problem classifier, use global...
The computation of indefinite integrals in some kinds “closed form”, the so-called symbolic integration, is an important and basic research subarea computer algebra. After implementing Risch’s algorithm partly, it was realized that efficient algorithms can be achieved parallel integration. One most famous Risch-Norman algorithm. However, this approach relies on analytic with heuristic degree bounds. Norman’s completion-based provides alternative for finding numerator integral avoiding...
In symbolic integration, the Risch--Norman algorithm aims to find closed forms of elementary integrals over differential fields by an ansatz for integral, which usually is based on heuristic degree bounds. Norman presented approach that avoids bounds and only relies completion reduction systems. We give a formalization his we develop refined process, terminates in more instances. some situations when does not terminate, one can detect patterns allowing still describe infinite systems are...
Despite recent breakthroughs in deep learning methods for image lighting enhancement, they are inferior when applied to portraits because 3D facial information is ignored their models. To address this, we present a novel framework portrait enhancement based on guidance. Our consists of two stages. In the first stage, corrected parameters predicted by network from input bad image, with assistance morphable model and differentiable renderer. Given parameter, renderer renders face shading...
Medical image segmentation is almost the most important pre-processing procedure in computer-aided diagnosis but also a very challenging task due to complex shapes of segments and various artifacts caused by medical imaging, (i.e., low-contrast tissues, non-homogenous textures). In this paper, we propose simple yet effective framework that incorporates geometric prior contrastive similarity into weakly-supervised loss-based fashion. The proposed built on point cloud provides meticulous...
Introduction: Breast cancer is the most common type of in women globally and mammograms are a primary method for diagnosing it. The challenge interpreting exacerbated by shortage dedicated breast radiologists lengthy training required to cultivate such expertise. This study designed investigate performance AI assistance resident compared consultant conduct cost analysis implementing diagnostic setting. Methods: A multi-reader multi-case was conducted at National University Hospital Singapore...