A5037049534- Dental Radiography and Imaging
- Medical Image Segmentation Techniques
- Image Processing Techniques and Applications
- Human Pose and Action Recognition
- Remote Sensing and Land Use
- Advanced Neural Network Applications
- Remote-Sensing Image Classification
- Image and Object Detection Techniques
- Infrared Target Detection Methodologies
- Advanced Image Fusion Techniques
- Advanced Fiber Laser Technologies
- Advanced Image and Video Retrieval Techniques
- 3D Shape Modeling and Analysis
- Adversarial Robustness in Machine Learning
- Brain Tumor Detection and Classification
- Video Surveillance and Tracking Methods
- Image Retrieval and Classification Techniques
- Anomaly Detection Techniques and Applications
- Medical Imaging Techniques and Applications
- Advanced Image Processing Techniques
- Advanced Fiber Optic Sensors
- Orthodontics and Dentofacial Orthopedics
- Biometric Identification and Security
- Medical Imaging and Analysis
- Domain Adaptation and Few-Shot Learning
Chongqing University of Posts and Telecommunications
2017-2025
Shandong University of Traditional Chinese Medicine
2025
Beijing Municipal Education Commission
2025
Beijing Forestry University
2020-2025
Research Institute of Forestry
2025
Zhejiang University
2024
Beijing Information Science & Technology University
2023-2024
Space Engineering University
2023-2024
Northwestern Polytechnical University
2022-2023
Harbin Institute of Technology
2023
Abstract Accurate delineation of individual teeth and alveolar bones from dental cone-beam CT (CBCT) images is an essential step in digital dentistry for precision healthcare. In this paper, we present AI system efficient, precise, fully automatic segmentation real-patient CBCT images. Our evaluated on the largest dataset so far, i.e., using a 4,215 patients (with 4,938 scans) 15 different centers. This achieves accuracy comparable to experienced radiologists (e.g., 0.5% improvement terms...
Existing studies of multi-modality medical image segmentation tend to aggregate all modalities without discrimination and employ multiple symmetric encoders or decoders for feature extraction fusion. They often overlook the different contributions visual representation intelligent decisions among images. Motivated by this discovery, paper proposes an asymmetric adaptive heterogeneous network with modality For extraction, it uses a two-stream feature-bridging extract complementary features...
Recently Adversarial Examples (AEs) that deceive deep learning models have been a topic of intense research interest. Compared with the AEs in digital space, physical adversarial attack is considered as more severe threat to applications like face recognition authentication, objection detection autonomous driving cars, etc. In particular, deceiving object detectors practically, challenging since relative position between and detector may keep changing. Existing works attacking are still very...
Document-level information is very important for event detection even at sentence level. In this paper, we propose a novel Document Embedding Enhanced Bi-RNN model, called DEEB-RNN, to detect events in sentences. This model first learns oriented embeddings of documents through hierarchical and supervised attention based RNN, which pays word-level triggers sentence-level those sentences containing events. It then uses the learned document embedding enhance another bidirectional RNN identify...
Cross-modal hashing that leverages hash functions to project high-dimensional data from different modalities into the compact common hamming space, has shown immeasurable potential in cross-modal retrieval. To ease labor costs, unsupervised methods are proposed. However, existing still suffer two factors optimization of functions: 1) similarity guidance, they barely give a clear definition whether is similar or not between points, leading residual redundant information; 2) strategy, ignore...
Infrared small target detection plays an important role in the infrared search and tracking applications. In recent years, deep learning techniques have been introduced to this task achieved noteworthy effects. Following general object segmentation methods, existing methods usually process image from global view. However, locality of targets extreme class-imbalance between background pixels are not well-considered by these which causes low-efficiency on training high-dependence numerous...
Training text-to-image models with web scale image-text pairs enables the generation of a wide range visual concepts from text. However, these pre-trained often face challenges when it comes to generating highly aesthetic images. This creates need for alignment post pre-training. In this paper, we propose quality-tuning effectively guide model exclusively generate visually appealing images, while maintaining generality across concepts. Our key insight is that supervised fine-tuning set...
This article proposes a robust technique for needle detection and tracking using three-dimensional ultrasound (3D US). It is difficult radiologists to detect follow the position of micro tools, such as biopsy needles, that are inserted in human tissues under 3D US guidance. To overcome this difficulty, we propose method automatically reduces processed volume limited region interest (ROI), increasing at same time calculation speed robustness proposed technique. First, line filter enhances...
Accurate tooth identification and delineation in dental CBCT images are essential clinical oral diagnosis treatment. Teeth positioned the alveolar bone a particular order, featuring similar appearances across adjacent bilaterally symmetric teeth. However, existing segmentation methods ignored such specific anatomical topology, which hampers accuracy. Here we propose semantic graph-based method to explicitly model spatial associations between different targets (i.e., teeth) for their precise...
Hyperspectral image (HSI) classification suffers from two serious problems, one is the limited labeled pixels, and other class imbalance problem. As a result, number of pixels in many categories not sufficient to characterize spectral-spatial information, train satisfying deep model. By making full use information unlabeled semi-supervised methods can provide better performance case pixels. However, they do take into account HSI data. method data enhancement, generative adversarial networks...
Evoked brain oscillations in the gamma range have been shown to assist stroke recovery. However, causal relationship between evoked and neuroprotection is not well understood. We used optogenetic stimulation investigate how modulate cortical dynamics acute phase after stroke. Our results reveal that at 40 Hz drives activity interneurons frequency phase-locked principal neurons a lower frequency, leading increased cross-frequency coupling. In addition, 40-Hz enhances interregional...
Due to the difficulty of manually labeling remote sensing scene images and demand for ability recognize new classes, few-shot classification (FSRSSC) has attracted more attention. At present, metric-based FSRSSC methods have made promising progress, especially prototypical networks-based methods. However, due complexity background images, prototype classifier, which takes average features support samples as metric benchmark, retains category-irrelevant objects other information in image....
Precise segmentation of teeth from intra-oral scanner images is an essential task in computer-aided orthodontic surgical planning. The state-of-the-art deep learning-based methods often simply concatenate the raw geometric attributes (i.e., coordinates and normal vectors) mesh cells to train a single-stream network for automatic image segmentation. However, since different reveal completely information, naive concatenation at (low-level) input stage may bring unnecessary confusion describing...
The ability to segment teeth precisely from digitized 3D dental models is an essential task in computer-aided orthodontic surgical planning. To date, deep learning based methods have been popularly used handle this task. State-of-the-art directly concatenate the raw attributes of inputs, namely coordinates and normal vectors mesh cells, train a single-stream network for fully-automated tooth segmentation. This, however, has drawback ignoring different geometric meanings provided by those...
Hyperspectral images (HSIs) not only possess abundant spectral features but also present a detailed spatial distribution of land cover, and they have significant advantages in the fine classification ground materials. Recently, using convolutional neural networks (CNNs) to extract spectral–spatial has become an effective way for HSI classification. However, conventional convolution kernels learn from fixed regular square regions, rich information been effectively explored. In this letter,...
Unmanned aerial vehicle (UAV) technology, artificial intelligence, and the relevant hardware can be used for monitoring wild animals. However, existing methods have several limitations. Therefore, this study explored protection of Amur tigers their main prey species using images from UAVs by optimizing algorithm models with respect to accuracy, model size, recognition speed, elimination environmental interference. Thermal imaging data were collected 2000 pictures a thermal lens on DJI...
Jiawei Kongsheng Zhenzhong Pill(JKZP) is based on Pill contained in the Tang Dynasty's "Thousand Golden Prescriptions," which exhibited good anti-ischemic and antidepressant effects previous study. However, its specific post-stroke depression (PSD) mechanism are not clear. This study aimed to investigate of JKZP treatment PSD related mechanisms. The decoction was first analyzed for medicinal chemical composition screened representative components JKZP. Middle cerebral artery occlusion (MCAO)...