A5060901632- Computational Geometry and Mesh Generation
- Advanced Neural Network Applications
- AI in cancer detection
- Medical Image Segmentation Techniques
- Robotic Path Planning Algorithms
- Radiomics and Machine Learning in Medical Imaging
- Data Management and Algorithms
- Advanced Radiotherapy Techniques
- Complexity and Algorithms in Graphs
- Medical Imaging Techniques and Applications
- Digital Image Processing Techniques
- Optimization and Search Problems
- Cell Image Analysis Techniques
- COVID-19 diagnosis using AI
- Advanced Graph Theory Research
- Advanced Numerical Analysis Techniques
- Computer Graphics and Visualization Techniques
- Digital Imaging for Blood Diseases
- Optimization and Packing Problems
- Domain Adaptation and Few-Shot Learning
- Radiation Therapy and Dosimetry
- Robotics and Sensor-Based Localization
- Medical Imaging and Analysis
- Algorithms and Data Compression
- Retinal Imaging and Analysis
University of Notre Dame
2016-2025
York Central Hospital
2019-2023
Hunan University
2022
Stony Brook University
2021
Cancer Research Institute
2020
Pfizer (United States)
2013-2020
Zhejiang University
2019
McGill University
2017
Northwestern University
2009
The Ohio State University
2003-2005
Segmentation of 3D images is a fundamental problem in biomedical image analysis. Deep learning (DL) approaches have achieved state-of-the-art segmentation perfor- mance. To exploit the contexts using neural networks, known DL methods, including convolution, 2D convolution on planes orthogonal to slices, and LSTM multiple directions, all suffer incompatibility with highly anisotropic dimensions common images. In this paper, we propose new framework for segmentation, based com- bination fully...
3D object detection on point clouds finds many applications. However, most known cloud methods did not adequately accommodate the characteristics (e.g., sparsity) of clouds, and thus some key semantic information shape information) is well captured. In this paper, we propose a new graph convolution (GConv) based hierarchical network (HGNet) for detection, which processes raw directly to predict bounding boxes. HGNet effectively captures relationship points utilizes multi-level semantics...
Abstract Early detection is critical to achieving improved treatment outcomes for child patients with congenital heart diseases (CHDs). Therefore, developing effective CHD techniques using low-cost and non-invasive pediatric electrocardiogram are highly desirable. We propose a deep learning approach detection, CHDdECG, which automatically extracts features from wavelet transformation characteristics, integrates them key human-concept features. Developed on 65,869 cases, CHDdECG achieved...
In this paper, we present a two-phase framework that integrates task assignment, ordering and voltage selection (VS) together to minimize energy consumption of real-time dependent tasks executing on given number variable processors. Task assignment in the first phase strive maximize opportunities can be exploited for lowering levels during second phase, i.e., selection. formulate VS problem as an Integer Programming (IP) solve IP efficiently. Experimental results demonstrate our is very...
With pervasive applications of medical imaging in health-care, biomedical image segmentation plays a central role quantitative analysis, clinical diagnosis, and intervention. Since manual annotation suffers limited reproducibility, arduous efforts, excessive time, automatic is desired to process increasingly larger scale histopathological data. Recently, deep neural networks (DNNs), particularly fully convolutional (FCNs), have been widely applied segmentation, attaining much improved...
Colorectal cancer (CRC) is one of the most life-threatening malignancies. Colonoscopy pathology examination can identify cells early-stage colon tumors in small tissue image slices. But, such time-consuming and exhausting on high resolution images. In this paper, we present a new framework for colonoscopy whole slide (WSI) analysis, including lesion segmentation diagnosis. Our contains an improved U-shape network with VGG net as backbone, two schemes training inference, respectively (the...
Many known supervised deep learning methods for medical image segmentation suffer an expensive burden of data annotation model training. Recently, few-shot were proposed to alleviate this burden, but such often showed poor adaptability the target tasks. By prudently introducing interactive into strategy, we develop a novel approach called Interactive Few-shot Learning (IFSL), which not only addresses models also tackles common issues methods. First, design new structure, Medical Prior-based...
The rib fracture is a common type of thoracic skeletal trauma, and its inspections using computed tomography (CT) scans are critical for clinical evaluation treatment planning. However, it often challenging radiologists to quickly accurately detect fractures due tiny objects blurriness in large 3D CT images. Previous diagnoses automatic mostly relied on deep learning (DL)-based object detection, which highly depends label quality quantity. Moreover, general detection methods did not take...
In this paper, we introduce U-Net v2, a new robust and efficient variant for medical image segmentation. It aims to augment the infusion of semantic information into low-level features while simultaneously refining high-level with finer details. For an input image, begin by extracting multi-level deep neural network encoder. Next, enhance feature map each level infusing from higher-level integrating details lower-level through Hadamard product. Our novel skip connections empower all levels...
Arc-modulated radiation therapy (AMRT) is a novel rotational intensity-modulated (IMRT) technique developed for clinical linear accelerator that aims to deliver highly conformal treatment using just one arc of gantry rotation. Compared fixed-gantry IMRT and the multiple-arc (IMAT) techniques, AMRT promises same quality with single-arc delivery. In this paper, we present planning scheme AMRT, which addresses challenges in inverse planning, leaf sequencing dose calculation. The feasibility...
Significance Microbial parasites may behave collectively to manipulate their host’s behavior. We examine adaptations of a microbial parasite in its natural environment: the body coevolved and manipulated host. Electron microscopy 3D reconstructions host tissues reveal that this fungus invades muscle fibers throughout ant’s but leaves brain intact, fungal cells connect form extensive networks. The connections are likened structures aid transporting nutrients organelles several...
Despite extensive interest, extracellular vesicle (EV) research remains technically challenging. One of the unexplored gaps in EV has been inability to characterize spatially and functionally heterogeneous populations EVs based on their metabolic profile. In this paper, we utilize intrinsic optical structural contrast demonstrate vivo/in situ characterization a variety unprocessed (pre)clinical samples. With pixel-level segmentation mask provided by deep neural network, individual can be...
Volumetric modulated arc therapy (VMAT) has found widespread clinical application in recent years. A large number of treatment planning studies have evaluated the potential for VMAT different disease sites based on currently available commercial implementations planning. In contrast, literature underlying mathematical optimization methods used is scarce. represents a challenging scale problem. contrast to fluence map intensity‐modulated radiotherapy static beams, nonconvex this paper,...