A5067387502- Medical Imaging and Analysis
- Advanced X-ray and CT Imaging
- Topic Modeling
- Radiomics and Machine Learning in Medical Imaging
- Radiation Dose and Imaging
- Natural Language Processing Techniques
- Multimodal Machine Learning Applications
- Spine and Intervertebral Disc Pathology
- Medical Imaging Techniques and Applications
- Emotion and Mood Recognition
- Advanced Computing and Algorithms
- Bone health and osteoporosis research
- Medical Image Segmentation Techniques
- Privacy-Preserving Technologies in Data
- Musculoskeletal synovial abnormalities and treatments
- Advanced MRI Techniques and Applications
- Gaze Tracking and Assistive Technology
- Spinal Fractures and Fixation Techniques
- EEG and Brain-Computer Interfaces
- Functional Brain Connectivity Studies
- Photoacoustic and Ultrasonic Imaging
- Misinformation and Its Impacts
- Cloud Data Security Solutions
- Complex Network Analysis Techniques
- Face and Expression Recognition
Peking University
2022-2025
Peking University Third Hospital
2022-2025
Shandong University of Traditional Chinese Medicine
2025
Suzhou Institute of Biomedical Engineering and Technology
2024-2025
Chinese Academy of Sciences
2024-2025
University of Hong Kong
2024
Tiangong University
2023
Background Pedicle screw loosening (PSL) is a frequent complication in osteoporotic patients undergoing spinal fixation, yet effective risk assessment methods are limited. This study explores the impact of craniocaudal cyclic load on pedicle fixation strength using computed tomography-based finite element analysis (CT-FEA) and evaluates its predictive value for PSL. Methods A total 23 PSL cases (7 men 16 women) 29 matched controls were analyzed CT-FEA. Both simple axial pullout with preset...
Against the background of globalization, circulation range traditional Chinese medicinal materials is constantly expanding, and phenomena mixed origins counterfeiting are becoming increasingly serious. Tracing origin great significance for ensuring their quality, safety, effectiveness. Laser-induced breakdown spectroscopy (LIBS), as a rapid non-destructive element analysis technique, can be used tracing materials. Deep learning not only handle non-linear relationships but also automatically...
5 T magnetic resonance imaging (MRI)-induced patient discomfort and the associated contributing factors remain unclear. To assess frequency of during MRI examinations analyze that may lead to discomfort, understand potential challenges, improve experience with systems. Prospective study. A total 539 participants, comprising patients healthy volunteers. 5.0 T. Each participant completed a post-MRI examination tolerance questionnaire evaluating overall experience. type was analyzed separately...
Multi-task model training has been adopted to enable a single deep neural network (often large language model) handle multiple tasks (e.g., question answering and text summarization). commonly receives input sequences of highly different lengths due the diverse contexts tasks. Padding (to same sequence length) or packing (short examples into long is usually prepare samples for training, which nonetheless not space computation efficient. This paper proposes dynamic micro-batching approach...
Abstract Objectives To evaluate image quality and diagnostic performance of carotid dual-energy computed tomography angiography (DECTA) using deep learning reconstruction (DLIR) compared with images adaptive statistical iterative reconstruction-Veo (ASIR-V). Methods Carotid DECTA datasets 28 consecutive patients were reconstructed at 50 keV DLIR low, medium, high levels (DLIR-L, DLIR-M, DLIR-H) 80% ASIR-V algorithms. Mean attenuation, noise, signal-to-noise ratio (SNR), contrast-to-noise...
Abstract Purpose MR arthrography (MRA) is the most accurate method for preoperatively diagnosing superior labrum anterior–posterior (SLAP) lesions, but diagnostic results can vary considerably due to factors such as experience. In this study, deep learning was used facilitate preliminary identification of SLAP lesions and compared with radiologists different seniority. Methods MRA data from 636 patients were retrospectively collected, all classified having/not having according shoulder...
Abstract Thus, the aim of this study is to evaluate performance deep learning imaging reconstruction (DLIR) algorithm in different image sets derived from carotid dual-energy computed tomography angiography (DECTA) for evaluating cervical intervertebral discs (IVDs) and compare them with those reconstructed using adaptive statistical iterative reconstruction-Veo (ASiR-V). Forty-two patients who underwent DECTA were included retrospective analysis. Three types (70 keV, water-iodine,...
Purpose This study aimed to evaluate the difference in vertebral mechanical properties estimated by finite element analysis (FEA) with different computed tomography (CT) reconstruction kernels and their accuracy screening classification of osteoporosis. Methods There were 31 patients enrolled retrospectively from quantitative CT database our hospital, uniformly covering range osteoporosis normal. All subjects’ raw data reconstructed both a smooth standard convolution kernel (B40f) sharpening...
<title>Abstract</title> Background: Screw loosening remains a prominent complication for osteoporotic patients with pedicle screw fixation surgeries, yet limited risk assessment approach. The aim of this study was to investigate influence craniocaudal cyclic load on strength by computed tomography (CT) based finite element analysis (FEA) and we examined predict ability in (PSL). Methods: 12 clinical PSL cases (7 men, 5 women) age- sex-matched controls were enrolled CT FEA. Simple axial...
The Mixture-of-Expert (MoE) technique plays a crucial role in expanding the size of DNN model parameters. However, it faces challenge extended all-to-all communication latency during training process. Existing methods attempt to mitigate this issue by overlapping with expert computation. Yet, these frequently fall short achieving sufficient overlap, consequently restricting potential for performance enhancements. In our study, we extend scope considering overlap at broader graph level....
Dual-energy computed tomography (DECT) has demonstrated the feasibility of using HAP-water to respond BMD changes without requiring dedicated software or calibration. Artificial intelligence (AI) been utilized for diagnosising osteoporosis in routine CT scans but rarely used DECT. This study investigated diagnostic performance an AI system screening DECT images with reference quantitative (QCT). prospective included 120 patients who underwent and QCT from August December 2023. Two...
Facial expression recognition (FER) is a challenging task. The following two significant problems often exist in real-life FER tasks: first, facial images the wild are uncertain, i.e., occlusion or blurred images; and second, there little intra-class similarity high inter-class similarity. In this paper, we propose Self-Curing network based on Coordinate attention Island loss(CoIsland-SCN) to address above problems. Firstly, coordinate module aggregates input features vertical horizontal...
With the rapid growth of Internet, people are exposed to a significant increase in amount information. However, dissemination false news has become rampant, causing adverse effects on society. Rumor detection emerged as an important and challenging research direction aimed at accurately determining authenticity messages. This paper proposes novel rumor method based knowledge graph attention networks. The leverages mechanism networks learn topic-enhanced representations extract features...
Multi-task model training has been adopted to enable a single deep neural network (often large language model) handle multiple tasks (e.g., question answering and text summarization). commonly receives input sequences of highly different lengths due the diverse contexts tasks. Padding (to same sequence length) or packing (short examples into long is usually prepare samples for training, which nonetheless not space computation efficient. This paper proposes dynamic micro-batching approach...
After years of development, deep convolutional neural networks have been widely used in the field polyp segmentation. In clinical medicine, segmentation polyps medical images plays an important role diagnosis and treatment diseases. However, due to different shapes sizes polyps, boundary between surrounding tissues mucosa is not obvious, it very difficult accurately segment polyps. Therefore, this paper introduces attention module reverse network improve accuracy At same time, a Channel-wise...