A5101740599- Domain Adaptation and Few-Shot Learning
- Medical Image Segmentation Techniques
- Advanced Neural Network Applications
- AI in cancer detection
- Fetal and Pediatric Neurological Disorders
- Advanced Image Processing Techniques
- Generative Adversarial Networks and Image Synthesis
- Advanced Vision and Imaging
- Robotics and Sensor-Based Localization
- Cleft Lip and Palate Research
- Image and Signal Denoising Methods
- Radiomics and Machine Learning in Medical Imaging
- Photoacoustic and Ultrasonic Imaging
- Advanced Image and Video Retrieval Techniques
- Hip disorders and treatments
- Image and Object Detection Techniques
- Face recognition and analysis
- Medical Imaging and Analysis
- COVID-19 diagnosis using AI
- Cardiovascular Health and Disease Prevention
- Cerebrovascular and Carotid Artery Diseases
- Visual Attention and Saliency Detection
- Digital Media Forensic Detection
- Brain Tumor Detection and Classification
- Handwritten Text Recognition Techniques
Shenzhen University
2020-2025
Shenzhen University Health Science Center
2020-2025
Third Affiliated Hospital of Sun Yat-sen University
2025
Sun Yat-sen University
2021-2025
Xi'an Jiaotong University
2018-2024
City University of Macau
2024
Xinjiang University
2023-2024
South China University of Technology
2021-2023
I-Shou University
2023
University of Macau
2022
The Segment Anything Model (SAM) is the first foundation model for general image segmentation. It has achieved impressive results on various natural segmentation tasks. However, medical (MIS) more challenging because of complex modalities, fine anatomical structures, uncertain and object boundaries, wide-range scales. To fully validate SAM's performance data, we collected sorted 53 open-source datasets built a large dataset with 18 84 objects, 125 object-modality paired targets, 1050K 2D...
Early diagnosis of cleft lip and palate (CLP) requires a multiplane examination, demanding high technical proficiency from radiologists. Therefore, this study aims to develop validate the first artificial intelligence (AI)-based model (CLP-Net) for fully automated multi-plane localization in three-dimensional(3D) ultrasound during trimester. This retrospective included 418 (394 normal, 24 CLP) 3D 288 pregnant woman between July 2022 October 2024 Shenzhen Guangming District People's Hospital...
In this paper, we present CLCC, a novel contrastive learning framework for color constancy. Contrastive has been applied high-quality visual representations image classification. One key aspect to yield useful classification is design illuminant invariant augmentations. However, the assumption conflicts with nature of constancy task, which aims estimate given raw image. Therefore, construct effective pairs better illuminant-dependent features via raw-domain augmentation. On NUS-8 dataset,...
The success of the state-of-the-art video deblurring methods stems mainly from implicit or explicit estimation alignment among adjacent frames for latent restoration. However, due to influence blur effect, estimating information blurry is not a trivial task. Inaccurate estimations will interfere following frame Instead information, we propose simple and effective deep Recurrent Neural Network with Multi-scale Bi-directional Propagation (RNN-MBP) effectively propagate gather unaligned...
This paper reviews the NTIRE 2022 Challenge on Super-Resolution and Quality Enhancement of Compressed Video. In this challenge, we proposed LDV 2.0 dataset, which includes dataset (240 videos) 95 additional videos. challenge three tracks. Track 1 aims at enhancing videos compressed by HEVC a fixed QP. 2 3 target both super-resolution quality enhancement video. They require x2 x4 super-resolution, respectively. The tracks totally attract more than 600 registrations. test phase, 8 teams, teams...
Efficient representation of point clouds is fundamental for LiDAR-based 3D object detection. While recent grid-based detectors often encode into either voxels or pillars, the distinctions between these approaches remain underexplored. In this paper, we quantify differences current encoding paradigms and highlight limited vertical learning within. To tackle limitations, propose a hybrid detection framework named Voxel-Pillar Fusion (VPF), which synergistically combines unique strengths both...
Deep Neural Networks (DNNs) suffer from the performance degradation when image appearance shift occurs, especially in ultrasound (US) segmentation. In this paper, we propose a novel and intuitive framework to remove shift, hence improve generalization ability of DNNs. Our work has three highlights. First, follow spirit universal style transfer shifts, which was not explored before for US images. Without sacrificing structure details, it enables arbitrary style-content transfer. Second,...
To correlate the molar teeth and their periodontal conditions after orthodontic anterior removal of molars to close gap missing provide risk factor analysis prognostic assessment for treatment such cases. Twenty-six patients (five males 21 females) with a total 42 were selected undergo full-mouth due absence first/second bone volume was recorded by cone-beam computed tomography (CBCT) scanning before (T0) (T1) treatment, root calculated three-dimensional reconstruction CBCT. Afterwards,...
<title>Abstract</title> Background Accurate fetal growth evaluation is crucial for monitoring health, with crown-rump length (CRL) being the gold standard estimating gestational age and assessing during first trimester. To enhance CRL accuracy efficiency, we developed an AI-based model (3DCRL-Net) using 3D U-Net architecture automatic landmark detection to achieve plane localization measurement in ultrasound. We then compared its performance that of experienced radiologists both 2D...
The traditional form of a water town is the result combination urban history, culture, and spatial structure. This study digitized historical maps ancient Wenzhou City from four developmental periods (1566, 1765, 1876, 1949); employed space syntax to compute integration, choice, intelligibility road networks; categorized annotated architecturally functional attributes city using geographic information systems software. findings indicate that (1) City’s morphological development gradually...
Accurate standard plane (SP) localization is the fundamental step for prenatal ultrasound (US) diagnosis. Typically, dozens of US SPs are collected to determine clinical 2D has perform scanning each SP, which time-consuming and operator-dependent. While 3D containing multiple in one shot inherent advantages less user-dependency more efficiency. Automatically locating SP very challenging due huge search space large fetal posture variations. Our previous study proposed a deep reinforcement...
This paper introduces a novel fault model, called the dual-cell-aware (DCA) which targets short defects locating between two adjacent standard cells placed in layout. A layout-based methodology is also presented to automatically extract valid DCA faults from targeted designs and cell libraries. The identified are outputted format that can be applied commercial ATPG tool for test generation. result of simulation based on industrial have demonstrated cannot fully covered by tests conventional...
Automatic and accurate detection of anatomical landmarks is an essential operation in medical image analysis with a multitude applications. Recent deep learning methods have improved results by directly encoding the appearance captured anatomy likelihood maps (i.e., heatmaps). However, most current solutions overlook another essence heatmap regression, objective metric for regressing target heatmaps rely on hand-crafted heuristics to set precision, thus being usually cumbersome...