A5030890710- Radiomics and Machine Learning in Medical Imaging
- Retinal Imaging and Analysis
- Advanced Neural Network Applications
- Brain Tumor Detection and Classification
- Medical Image Segmentation Techniques
- AI in cancer detection
- Generative Adversarial Networks and Image Synthesis
- Digital Imaging for Blood Diseases
- Retinal Diseases and Treatments
- Retinal and Optic Conditions
- Face recognition and analysis
- Anomaly Detection Techniques and Applications
- Advanced Image Fusion Techniques
- COVID-19 diagnosis using AI
- Advanced Multi-Objective Optimization Algorithms
- Image and Signal Denoising Methods
- Face and Expression Recognition
- Metaheuristic Optimization Algorithms Research
- Advanced Image Processing Techniques
- Imbalanced Data Classification Techniques
- Advanced Vision and Imaging
- Biometric Identification and Security
- 3D Shape Modeling and Analysis
- Ophthalmology and Visual Impairment Studies
- Adversarial Robustness in Machine Learning
Fudan University
2020-2024
Huashan Hospital
2021-2024
Shanghai Chest Hospital
2024
Chinese Academy of Sciences
2024
Beijing University of Posts and Telecommunications
2023
Beijing Chao-Yang Hospital
2023
Capital Medical University
2023
Syracuse University
2023
Anhui University of Science and Technology
2023
Tsinghua University
2023
Medical datasets, especially medical images, are often imbalanced due to the different incidences of various diseases. To address this problem, many methods have been proposed synthesize images using generative adversarial networks (GANs) enlarge training datasets for facilitating image analysis. For instance, conventional such as image-to-image translation techniques used fundus with their respective vessel trees in field image.In order improve quality and details synthetic three key...
Abstract Background Chest CT is used for the assessment of severity patients infected with novel coronavirus 2019 (COVID-19). We collected chest scans 202 diagnosed COVID-19, and try to develop a rapid, accurate automatic tool screening follow-up therapeutic treatment. Methods A total 729 2D axial plan slices 246 severe cases 483 non-severe were employed in this study. By taking advantages pre-trained deep neural network, four off-the-shelf models (Inception-V3, ResNet-50, ResNet-101,...
Microaneurysms (MAs) play an important role in the diagnosis of clinical diabetic retinopathy at early stage. Annotation MAs manually by experts is laborious and so it essential to develop automatic segmentation methods. Automatic MA remains a challenging task mainly due low local contrast image small size MAs. A deep learning-based method called U-Net has become one most popular methods for medical task. We propose architecture U-Net, named recurrent (DRU-Net), obtained combining residual...
In this paper, we propose a progressive margin loss (PML) approach for unconstrained facial age classification. Conventional methods make strong assumption on that each class owns adequate instances to outline its data distribution, likely leading bias prediction where the training samples are sparse across classes. Instead, our PML aims adaptively refine label pattern by enforcing couple of margins, which fully takes in in-between discrepancy intra-class variance, inter-class variance and...
Background Differential diagnosis of primary central nervous system lymphoma (PCNSL) and glioblastoma (GBM) is useful to guide treatment strategies. Purpose To investigate the use a convolutional neural network (CNN) model for differentiation PCNSL GBM without tumor delineation. Study Type Retrospective. Population A total 289 patients with (136) or (153) were included, average age cohort was 54 years, there 173 men 116 women. Field Strength/Sequence 3.0 T Axial contrast‐enhanced 1 ‐weighted...
Optical coherence tomography angiography (OCTA) is a relatively new imaging modality that generates microvasculature map. Meanwhile, deep learning has been recently attracting considerable attention in image-to-image translation, such as image denoising, super-resolution and prediction. In this paper, we propose based pipeline for OCTA. This consists of three parts: training data preparation, model OCTA predicting using the trained model. To be mentioned, datasets used work were...
Diabetic retinopathy (DR) is a leading cause of visual blindness. However if DR can be diagnosed and treated early, 90% causing blindness prevented significantly. Microaneurysms (MAs) exudates (EXs), as signs DR, used for early diagnosis. However, MAs EXs segmentation challenging task due to the low contrast lesions, interference noises, imbalance between lesion areas background. In this paper, an enhanced residual U-Net (ERU-Net) proposed. ERU-Net obtains three U-paths, which are composed...
Human activity recognition is an imbalance classification problem in essence since various human actions may occur at different frequencies. Traditional ensemble class learning methods integrate resampling technique with multi-classifier models, which obtain better generalization than single ones. However, the number of base classifiers often determined advance, resulting redundant structure a and larger computational cost. Moreover, combinations have similar accuracy, forming multimodel...
Abstract Purpose The objective of this study is to construct a computer aided diagnosis system for normal people and pneumoconiosis using X-raysand deep learning algorithms. Materials methods 1760 anonymous digital X-ray images real patients between January 2017 June 2020 were collected experiment. In order concentrate the feature extraction ability model more on lung region restrain influence external background factors, two-stage pipeline from coarse fine was established. First, U-Net used...
In the past few years, U-Net based U-shaped architecture and skip-connections have made incredible progress in field of medical image segmentation. U
Axial myopia is the most common type of myopia. However, due to high incidence in Chinese children, few studies estimating physiological elongation ocular axial length (AL), which does not cause progression and differs from non-physiological AL, have been conducted. The purpose our study was construct a machine learning (ML)-based model for AL sample school-aged myopic children.In total, 1011 children aged 6 18 years participated this study. Cross-sectional datasets were used optimize ML...
Neural architecture search (NAS) has made incredible progress in medical image segmentation tasks, due to its automatic design of the model. However, spaces studied many existing studies are based on U-Net and variants, which limits potential neural modeling better architectures. In this study, we propose a new NAS named GNAS-U<sup>2</sup>Net for joint optic cup disc. This is first application two-level nested U-shaped structure. The best performance achieved by model designed REFUGE dataset...
With the widespread popularity of electronic devices, emergence biometric technology has brought significant convenience to user authentication compared with traditional password and mode unlocking. Among many biological characteristics, face is a universal irreplaceable feature simple detection methods good recognition accuracy. Face one main functions equipment propaganda. The previous work in this field mainly focused on converting loss function deep convolution neural networks without...