A5100438471- Advanced MRI Techniques and Applications
- Medical Imaging Techniques and Applications
- Radiomics and Machine Learning in Medical Imaging
- Parallel Computing and Optimization Techniques
- Medical Image Segmentation Techniques
- AI in cancer detection
- Low-power high-performance VLSI design
- Advanced Neural Network Applications
- Advanced Neuroimaging Techniques and Applications
- Manufacturing Process and Optimization
- Image Processing Techniques and Applications
- Optical Imaging and Spectroscopy Techniques
- Advanced X-ray and CT Imaging
- Industrial Vision Systems and Defect Detection
- Advancements in Photolithography Techniques
- Digital Imaging for Blood Diseases
- Photoacoustic and Ultrasonic Imaging
- Cardiac Imaging and Diagnostics
- Non-Destructive Testing Techniques
- Sparse and Compressive Sensing Techniques
- Functional Brain Connectivity Studies
- Numerical Methods and Algorithms
- Advanced machining processes and optimization
- Herpesvirus Infections and Treatments
- COVID-19 diagnosis using AI
Weill Cornell Medicine
2021-2024
Cornell University
2018-2024
Northwestern Polytechnical University
2022-2024
Nanjing Normal University
2024
Nanjing Medical University
2022-2024
Alibaba Group (Cayman Islands)
2024
Ministry of Industry and Information Technology
2022
Wuhan University of Technology
2022
Wuhan University of Science and Technology
2022
Sun Yat-sen University
2022
While high-level synthesis (HLS) offers sophisticated techniques to optimize designs for area and performance, HLS-estimated resource usage timing often deviate significantly from actual quality of results (QoR) achieved by FPGA-targeted designs. Inaccurate HLS estimates prevent designers performing meaningful design space exploration without resorting the time-consuming downstream implementation process. To address this challenge, we first build a large collection C-to-FPGA diverse set...
With the continuous shrinking of technology nodes, lithography hotspot detection and elimination in physical verification phase is great value. Recently machine learning pattern matching based methods have been extensively studied to overcome runtime overhead problem expensive full-chip simulation. However, there still much room for improvement terms accuracy Overall Detection Simulation Time (ODST). In this paper, we propose a unified framework, where feature extraction optimization guided...
Deep reinforcement learning (DRL) has recently shown its success in tackling complex combinatorial optimization problems. When these problems are extended to multiobjective ones, it becomes difficult for the existing DRL approaches flexibly and efficiently deal with multiple subproblems determined by weight decomposition of objectives. This article proposes a concise meta-learning-based approach. It first trains meta-model meta-learning. The is fine-tuned few update steps derive submodels...
Deep learning (DL) is increasingly used to solve ill-posed inverse problems in medical imaging, such as reconstruction from noisy and/or incomplete data, DL offers advantages over conventional methods that rely on explicit image features and hand engineered priors. However, supervised DL-based may achieve poor performance when the test data deviates training for example, it has pathologies not encountered data. Furthermore, reconstructions do always incorporate underlying forward physical...
World Health Organization (WHO) group 1 pulmonary arterial hypertension (PAH) is a progressive, debilitating disease. Previous observational studies have demonstrated that artery denervation (PADN) reduces pressures in PAH. However, the safety and effectiveness of PADN not been established randomized trial. The aim this study was to determine treatment effects patients with Patients WHO PAH taking PAH-specific drugs for at least 30 days were enrolled multicenter, sham-controlled,...
Chronic active multiple sclerosis (MS) lesions are characterized by a paramagnetic rim at the edge of lesion and associated with increased disability in patients. Quantitative susceptibility mapping (QSM) is an MRI technique that sensitive to chronic lesions, termed + on QSM. We present QSMRim-Net, data imbalance-aware deep neural network fuses lesion-level radiomic convolutional image features for automated identification
As the complexity and scale of FPGA circuits grows, resolving routing congestion becomes more important in placement. In this paper, we propose a routability-driven placement algorithm for large-scale heterogeneous FPGAs. Our proposed consists (1) partitioning, (2) packing, (3) global with estimation, (4) window-base legalization, (5) resource-aware detailed Experimental results show that our approach can give routable all benchmarks ISPD2016 contest achieve good result compared to other...
Accurate detection and segmentation of multiple sclerosis (MS) brain lesions on magnetic resonance images are important for disease diagnosis treatment. This is a challenging task as vary greatly in size, shape, location, image contrast. The objective our study was to develop an algorithm based deep convolutional neural network integrated with anatomic information lesion-wise loss function (ALL-Net) fast accurate automated MS lesions. Distance transformation mapping used construct module...
To improve accuracy and speed of quantitative susceptibility mapping plus blood oxygen level-dependent magnitude (QSM+qBOLD or QQ) -based extraction fraction (OEF) using a deep neural network (QQ-NET).The 3D multi-echo gradient echo images were acquired in 34 ischemic stroke patients 4 healthy subjects. Arterial spin labeling diffusion weighted imaging (DWI) also performed the patients. NET was developed to solve QQ model inversion problem based on Unet. QQ-based OEF maps reconstructed with...
Aqueous solubility is one of the most important physicochemical properties in drug discovery. At present, prediction aqueous compounds still a challenging problem. Machine learning has shown great potential prediction. Most machine models largely rely on setting hyperparameters, and their performance can be improved by hyperparameters better way. In this paper, we used MACCS fingerprints to represent structural features optimized light gradient boosting (LightGBM) with cuckoo search...
Quantitative susceptibility mapping (QSM) involves acquisition and reconstruction of a series images at multi-echo time points to estimate tissue field, which prolongs scan requires specific technique. In this paper, we present our new framework, called Learned Acquisition Reconstruction Optimization (LARO), aims accelerate the gradient echo (mGRE) pulse sequence for QSM. Our approach optimizing Cartesian k-space sampling pattern with deep network. Next, optimized was implemented in an mGRE...
Floating point multiplication is one of the most frequently used arithmetic operations in a wide variety applications, but high power consumption IEEE-754 standard floating multiplier prohibits its implementation many low systems, such as wireless sensors and other battery-powered embedded limits performance scaling CPUs GPGPUs for scientific computation. This paper presents low-power accuracy-configurable based on Mitchell's Algorithm. Post-layout SPICE simulations 45nm process show...
Continuous shrinking of VLSI technology nodes brings us powerful chips with lower power consumption, but it also introduces many issues in manufacturability. Lithography simulation process for new feature size suffers from large computational overhead. As a result, conventional mask optimization has been drastically resource consuming terms both time and cost. In this paper, we propose high performance machine learning-based printability evaluation framework lithography-related applications,...
Optical proximity correction (OPC) for advanced technology node now has become extremely expensive and challenging. Conventional model-based OPC encounters performance degradation large process variation, while aggressive approach such as inverse lithography (ILT) suffers from computational overhead both mask optimization writing processes. In this paper, we developed Neural-ILT, an end-to-end learning-based framework, which literally conducts prediction ILT a given layout in single neural...
Multiple sclerosis (MS) lesions occupy a small fraction of the brain volume, and are heterogeneous with regards to shape, size locations, which poses great challenge for training deep learning based segmentation models. We proposed new geometric loss formula address data imbalance exploit property MS lesions. showed that traditional region-based boundary-aware functions can be associated formula. further develop instantiate two containing first- second-order information lesion regions...
Advanced semiconductor process technologies are producing various circuit layout patterns, and it is essential to detect eliminate problematic ones, which called lithography hotspots. These hotspots formed due light diffraction interference, induces complex intrinsic structures within the formation process. Though machine learning based methods have been proposed for this problem, most of them cannot capture structure each data. In paper, we propose a novel feature extraction by representing...
We worked with Nestle SHIELD (Skin Health, Innovation, Education, and Longevity Development, NSH) to develop a deep learning model that is able assess acne severity from selfie images as accurate dermatologists. The was deployed mobile application, providing patients an easy way track the progress of their treatment. NSH acquired 4,700 for this study recruited 11 internal dermatologists label them in five categories: 1-Clear, 2- Almost Clear, 3-Mild, 4-Moderate, 5-Severe. Using OpenCV detect...
Medical images are often characterized by their structured anatomical representations and spatially inhomogeneous contrasts. Leveraging priors in neural networks can greatly enhance utility resource-constrained clinical settings. Prior research has harnessed such information for image segmentation, yet progress deformable registration been modest. Our work introduces textSCF, a novel method that integrates covariant filters textual prompts encoded visual-language models, to fill this gap....