A5041502507- Medical Imaging Techniques and Applications
- Advanced X-ray and CT Imaging
- Radiation Dose and Imaging
- AI in cancer detection
- Radiomics and Machine Learning in Medical Imaging
- Advanced MRI Techniques and Applications
- Image and Signal Denoising Methods
- Medical Image Segmentation Techniques
- Traffic Prediction and Management Techniques
- Digital Radiography and Breast Imaging
- Anomaly Detection Techniques and Applications
- Traffic control and management
- Smart Grid and Power Systems
- Ultrasound Imaging and Elastography
- Advanced Decision-Making Techniques
- Electrical and Bioimpedance Tomography
- COVID-19 Impact on Reproduction
- Smart Grid Security and Resilience
- Aerodynamics and Fluid Dynamics Research
- Long-Term Effects of COVID-19
- Opinion Dynamics and Social Influence
- Advanced Algorithms and Applications
- COVID-19 diagnosis using AI
- COVID-19 Clinical Research Studies
- Advanced Image Processing Techniques
Rensselaer Polytechnic Institute
2024-2025
Imaging Center
2025
Stony Brook University
2019-2024
State University of New York
2019-2024
Xi'an Jiaotong University
2016-2021
Soochow University
2021
Harbin Institute of Technology
2012-2015
Heilongjiang Institute of Technology
2012-2013
Aviation Industry Corporation of China (China)
2007
The continuous development and extensive use of computed tomography (CT) in medical practice has raised a public concern over the associated radiation dose to patient. Reducing may lead increased noise artifacts, which can adversely affect radiologists' judgment confidence. Hence, advanced image reconstruction from low-dose CT data is needed improve diagnostic performance, challenging problem due its ill-posed nature. Over past years, various methods have produced impressive results....
<title>Abstract</title> Recently, topological deep learning (TDL), which integrates algebraic topology with neural networks, has achieved tremendous success in processing point-cloud data, emerging as a promising paradigm data science. However, TDL not been developed for on differentiable manifolds, including images, due to the challenges posed by differential topology. We address this challenge introducing manifold (MTDL) first time. To highlight power of Hodge theory rooted topology, we...
As a follow-up to the first IEEE Transactions on Medical Imaging (TMI) special issue theme of deep tomographic reconstruction, second is assembled reflect status and momentum this rapidly emerging field. In editorial, we provide brief background illustrating motivation for development network-based, data-driven, learning-oriented reconstruction methods, summarize included papers, report our verification shared learning codes. Finally, discuss several important research topics facilitate...
Photon-counting spectral computed tomography (CT) is capable of material characterization and can improve diagnostic performance over traditional clinical CT. However, it suffers from photon count starving for each individual energy channel which may cause severe artifacts in the reconstructed images. Furthermore, since images different channels describe same object, there are high correlations among channels. To make full use inter-channel minimize effect while maintaining clinically...
BACKGROUND: In recent years, deep reinforcement learning (RL) has been applied to various medical tasks and produced encouraging results. OBJECTIVE: this paper, we demonstrate the feasibility of RL for denoising simulated deep-silicon photon-counting CT (PCCT) data in both full interior scan modes. PCCT offers higher spatial spectral resolution than conventional CT, requiring advanced methods suppress noise increase. METHODS: work, apply a dueling double Q network (DDDQN) denoise maximum...
Photon-counting computed tomography (PCCT) reconstructs multiple energy-channel images to describe the same object, where there exists a strong correlation among different channel images. In addition, reconstruction of each image suffers photon count starving problem. To make full use suppress data noise and enhance texture details in reconstructing image, this paper proposes tensor neural network (TNN) architecture learn multi-channel prior for PCCT reconstruction. Specifically, we first...
Lowering radiation dose per view and utilizing sparse views scan are two common CT modes, albeit often leading to distorted images characterized by noise streak artifacts. Blind image quality assessment (BIQA) strives evaluate perceptual in alignment with what radiologists perceive, which plays an important role advancing low-dose reconstruction techniques. An intriguing direction involves developing BIQA methods that mimic the operational characteristic of human visual system (HVS). The...
Deep learning methods have impacted almost every research field, demonstrating notable successes in medical imaging tasks such as denoising and super-resolution. However, the prerequisite for deep is data at scale, but sharing expensive yet risk of privacy leakage. As cutting-edge AI generative models, diffusion models now become dominant because their rigorous foundation unprecedented outcomes. Here we propose a latent approach synthesis without compromising patient privacy. In our...
Most penalized maximum likelihood methods for tomographic image reconstruction based on Bayes’ law include a freely adjustable hyperparameter to balance the data fidelity term and prior/penalty specific noise–resolution tradeoff. The is determined empirically via trial-and-error fashion in many applications, which then selects optimal result from multiple iterative reconstructions. These are not only time-consuming by their nature, but also require manual adjustment. This study aims...
Purpose: Bayesian theory provides a sound framework for ultralow-dose computed tomography (ULdCT) image reconstruction with two terms modeling the data statistical property and incorporating priori knowledge that is to be reconstructed. We investigate feasibility of using machine learning (ML) strategy, particularly convolutional neural network (CNN), construct tissue-specific texture prior from previous full-dose tomography. Approach: Our study constructs four priors, corresponding lung,...
Bayesian theory lies down a sound framework for ultralow-dose computed tomography (ULdCT) image reconstruction with two terms modeling the data statistical property and incorporating priori knowledge tobe- reconstructed image. This study investigates feasibility of using machine learning strategy, particularly convolutional neural network (CNN), to construct tissue-specific texture prior from previous full-dose CT (FdCT) integrates pre-log shift Poisson (SP) ULdCT images. The was implemented...
Abstract Tissue texture reflects the spatial distribution of contrasts image voxel gray levels, i.e., tissue heterogeneity, and has been recognized as important biomarkers in various clinical tasks. Spectral computed tomography (CT) is believed to be able enrich by providing different contrast images using X-ray energies. Therefore, this paper aims address two related issues for usage spectral CT, especially photon counting CT (PCCT): (1) enhancement reconstruction, (2) energy enriched...
CT is a main modality for imaging liver diseases, valuable in detecting and localizing tumors. Traditional anomaly detection methods analyze reconstructed images to identify pathological structures. However, these may produce suboptimal results, overlooking subtle differences among various tissue types. To address this challenge, here we employ generative diffusion prior inpaint the as reference facilitating detection. Specifically, use an adaptive threshold extract mask of abnormal regions,...
The elasticity of soft tissues has been widely considered a characteristic property for differentiation healthy and lesions and, therefore, motivated the development several imaging modalities, example, ultrasound elastography, magnetic resonance optical coherence elastography to directly measure tissue elasticity. This paper proposes an alternative approach modeling prior knowledge-based extraction elastic features machine learning (ML) lesion classification using computed tomography (CT)...
Interior photon-counting computed tomography (PCCT) scans are essential for obtaining high-resolution images at minimal radiation dose by focusing only on a region of interest. However, designing deep learning model denoising PCCT interior scan is rather challenging. Recently, several studies explored reinforcement (RL)-based models with far fewer parameters than those typical supervised and self-learning models. Such an RL can be effectively trained small dataset, yet generalizable...
Computed tomography (CT) involves a patient's exposure to ionizing radiation. To reduce the radiation dose, we can either lower X-ray photon count or down-sample projection views. However, of ways often compromises image quality. address this challenge, here introduce an iterative reconstruction algorithm regularized by diffusion prior. Drawing on exceptional imaging prowess denoising probabilistic model (DDPM), merge it with procedure that prioritizes data fidelity. This fusion capitalizes...
The elasticity of soft tissues has been widely considered as a characteristic property to differentiate between healthy and vicious and, therefore, motivated several imaging modalities, such Ultrasound Elastography, Magnetic Resonance Optical Coherence Elastography. This paper proposes an alternative approach modeling the using Computed Tomography (CT) modality for model-based feature extraction machine learning (ML) differentiation lesions. model describes dynamic non-rigid (or elastic)...
Spectral CT works with multiple energy X-ray sources or recognized detector, which can obtain the attenuation map under different energy. Therefore, spectral enhance contrast between soft tissues and enrich textures of colon polyps. In this paper, inspired by CT, we proposed a novel data engineering method, could prominently polyp classification enriched tissue textures. 63 volumetric images (31 benign 32 malignant) were obtained from clinical scanner effective 75keV All polyps resected...
Machine learning, especially convolutional neural network (CNN) approach has been successfully applied in noise suppression natural image. However, shifting from image to medical filed remains challenging due specific difficulties such as training samples limitation, clinically meaningful quality requirement and so on. To address this challenge, one possible solution is incorporate our human prior knowledge into the machine learning model better benefit its power. Therefore, work, we propose...
As an important part of the smart grid, substation is key operating parameters collection point and control execution point, thus its safe stable operation one fundamental bases ensuring grid continues to provide a reliable energy resource for development national economy.Using more new information communication technologies, which greatly improve efficiency intelligence, typical cyber-physical system (CPS), facing various security risks.Different from traditional view cyber power safety,...