A5025487788- COVID-19 diagnosis using AI
- Anomaly Detection Techniques and Applications
- AI in cancer detection
- Human Pose and Action Recognition
- Cell Image Analysis Techniques
- Hand Gesture Recognition Systems
- Medical Imaging Techniques and Applications
- Domain Adaptation and Few-Shot Learning
- Digital Media Forensic Detection
- Medical Image Segmentation Techniques
- Face and Expression Recognition
- Cerebrovascular and Carotid Artery Diseases
- Video Surveillance and Tracking Methods
- 3D Surveying and Cultural Heritage
- Computer Graphics and Visualization Techniques
- Retinal Imaging and Analysis
- Data-Driven Disease Surveillance
- Image Processing Techniques and Applications
- Advanced Image and Video Retrieval Techniques
- Face recognition and analysis
- Advanced Vision and Imaging
- Advanced X-ray and CT Imaging
- Machine Learning in Healthcare
- Advanced Image Processing Techniques
- Radiomics and Machine Learning in Medical Imaging
Emory University
2018-2023
Jilin Medical University
2023
Jilin University
2023
University of Chinese Academy of Sciences
2016-2018
In this article, we focus on isolated gesture recognition and explore different modalities by involving RGB stream, depth saliency stream for inspection. Our goal is to push the boundary of realm even further proposing a unified framework that exploits advantages multi-modality fusion. Specifically, spatial-temporal network architecture based consensus-voting has been proposed explicitly model long-term structure video sequence reduce estimation variance when confronted with comprehensive...
Recently, the popularity of depth-sensors such as Kinect has made depth videos easily available while its advantages have not been fully exploited. This paper investigates, for gesture recognition, to explore spatial and temporal information complementarily embedded in RGB sequences. We propose a convolutional twostream consensus voting network (2SCVN) which explicitly models both short-term long-term structure To alleviate distractions from background, 3d depth-saliency ConvNet stream...
Measurements of breast arterial calcifications (BAC) can offer a personalized, non-invasive approach to risk-stratify women for cardiovascular diseases such as heart attack and stroke. We aim detect segment in mammograms accurately suggest novel measurements quantify detected BAC future clinical applications.To separate mammograms, we propose lightweight fine vessel segmentation method Simple Context U-Net (SCU-Net). Due the large image size adopt patch-based way train SCU-Net obtain final...
Purpose: Existing anomaly detection methods focus on detecting interclass variations while medical image novelty identification is more challenging in the presence of intraclass variations. For example, a model trained with normal chest x-ray and common lung abnormalities expected to discover flag idiopathic pulmonary fibrosis, which rare disease unseen during training. The nuances lack relevant training data analysis pose great challenges for existing methods. Approach: We address above by...
Tumor-infiltrating lymphocytes (TILs) act as immune cells against cancer tissues. The manual assessment of TILs is usually erroneous, tedious, costly and subject to inter- intraobserver variability. Machine learning approaches can solve these issues, but they require a large amount labeled data for model training, which expensive not readily available. In this study, we present an efficient generative adversarial network,
Highly clumped nuclei captured in fluorescence microscopy images are commonly observed a wide spectrum of tissue-related biomedical investigations. To ensure the quality downstream analyses, it is essential to accurately segment clustered nuclei. However, this presents technical challenge as intensity alone often insufficient for recovering true boundaries. In paper, we propose an segmentation algorithm that identifies point pair connection candidates and evaluates adjacent connections with...
Abstract Motivation Morphological analyses with flatmount fluorescent images are essential to retinal pigment epithelial (RPE) aging studies and thus require accurate RPE cell segmentation. Although rapid technology advances in deep learning semantic segmentation have achieved great success many biomedical research, the performance of these supervised methods for is still limited by inadequate training data high-quality annotations. Results To address this problem, we develop a...
Highly clumped nuclei clusters captured in fluorescence situ hybridization microscopy images are common histology entities under investigations a wide spectrum of tissue-related biomedical investigations. Due to their large scale presence, computer based image analysis is used facilitate such with improved efficiency and reproducibility. To ensure the quality downstream analyses, it essential segment clustered high quality. However, this presents technical challenge commonly encountered...
Detecting out-of-distribution (OOD) samples in medical imaging plays an important role for downstream diagnosis. However, existing OOD detectors are demonstrated on natural images composed of inter-classes and have difficulty generalizing to images. The key issue is the granularity data domain, where intra-class predominant. We focus generalizability detection propose a self-supervised Cascade Variational autoencoder-based Anomaly Detector (CVAD). use variational autoencoders' cascade...
Detecting out-of-distribution samples for image applications plays an important role in safeguarding the reliability of machine learning model deployment. In this article, we developed a software tool to support our OOD detector CVAD - self-supervised Cascade Variational autoencoder-based Anomaly Detector , which can be easily applied various without any assumptions. The corresponding open-source is published better public research and usage.
Improving the retrieval relevance on noisy datasets is an emerging need for curation of a large-scale clean dataset in medical domain. While existing methods can be applied class-wise (aka. inter-class), they cannot distinguish granularity likeness within same class intra-class). The problem exacerbated external datasets, where samples are treated equally during training. Our goal to identify both intra/inter-class similarities fine-grained retrieval. To achieve this, we propose...
Language modality within the vision language pre-training framework is innately discretized, endowing each word in vocabulary a semantic meaning. In contrast, visual inherently continuous and high-dimensional, which potentially prohibits alignment as well fusion between modalities. We therefore propose to "discretize" representation by joint learning codebook that imbues token semantic. then utilize these discretized semantics self-supervised ground-truths for building our Masked Image...
The texture boundaries normally have frequent co-occurrence in natural images, but existing image enhancement techniques for the most parts focus on sharpening edges, i.e., intensity discontinuities. Moreover, these approaches often suffer from noise over-emphasis and extra artifact production. In this paper, we propose an adaptive boundary boosting algorithm. proposal exploits filter dual-layer decomposition pixel-wise amplification factor calculation synthesis. This leads to a...