Deep neural networks have demonstrated impressive results in medical image analysis, but designing suitable architectures for each specific task is expertise-dependent and time-consuming. Neural architecture search (NAS) offers an effective means of discovering architectures. It has been highly successful numerous applications, particularly natural classification. Yet, images possess unique characteristics, such as small regions a wide variety lesion sizes, that differentiate them from...

Detection of atrial fibrillation is important for risk stratification stroke. We developed a novel methodology to classify electrocardiograms (ECGs) normal, and other cardiac dysrhythmias as defined by the PhysioNet Challenge 2017.More specifically, we used piecewise linear splines feature selection gradient boosting algorithm classifier. In algorithm, ECG waveform fitted spline, morphological features relating spline coefficients are extracted. XGBoost heart rate variability features.The...

Many scientific studies collect data where the response and predictor variables are both functions of time, location, or some other covariate. Understanding relationship between these functional is a common goal in studies. Motivated from two real-life examples, we present this paper function-on-function regression model that can be used to analyze such kind data. Our estimator 2D coefficient function optimizer form penalized least squares penalty enforces certain level smoothness on...

Intrinsic bioinertness severely hampers the application of polyetheretherketone (PEEK), although in field dentistry it is considered to be an ideal titanium substitute implanting material. In this study, a bioactive silicate coating was successfully introduced onto PEEK surface by using electron beam evaporation (EBE) technology improve its bioactivity and osseointegration PEEK. Through controlling duration EBE, incorporated amounts silicon (Si) could exquisitely adjusted obtain proper...

Generative adversarial networks (GANs) are a powerful generative technique but frequently face challenges with training stability. Network architecture plays significant role in determining the final output of GANs, designing fine demands extensive domain expertise. This paper aims to address this issue by searching for high-performance generator's architectures through neural search (NAS). The proposed approach, called evolutionary weight sharing (EWSGAN), is based on and comprises two...

This Special Report summarizes the 2022, AAPM grand challenge on Truth-based CT image reconstruction. To provide an objective framework for evaluating reconstruction methods using virtual imaging resources consisting of a library simulated projection images population human models with various diseases. Two hundred unique anthropomorphic, computational were created varied diseases 67 emphysema, lung lesions, and 66 liver lesions. The organs modeled based clinical real patients. emphysematous...

Aiming at the problems of limited number electromagnetic real data and insufficient features in radar target intelligent identification, a generation model based on image mask constraints is proposed to realize effective expansion Range Doppler (R-D) images. First, set 1-D echo non-cooperative ship targets under sea conditions acquired. The velocity information well preserved by using Fourier transform signal preprocessing stage, interactive program designed perform annotation form an...

The detection of sea-surface ships represents a crucial means by which countries can compete for marine resources in the context current international strategic environment. ongoing advancement deep learning has led to notable enhancement accuracy target tasks involving surface ships, particularly when these are conducted using intelligent algorithms. Concurrently, depth and width network models expanding, accompanied an increase complexity. This resulted challenge deploying that limited...

Computed Tomography (CT) Image Reconstruction is an important technique used in a wide range of applications, ranging from explosive detection, medical imaging to scientific imaging. Among available reconstruction methods, Model Based Iterative (MBIR) produces higher quality images and allows for the use more general CT scanner geometries than possible with commonly methods. The high computational cost MBIR, however, often makes it impractical applications which would otherwise be ideal....

Large language models (LLMs) have demonstrated remarkable success as foundational models, benefiting various downstream applications through fine-tuning. Loss scaling studies the superior performance of larger LLMs compared to their smaller counterparts. Nevertheless, training with billions parameters poses significant challenges and requires considerable computational resources. For example, a one trillion parameter GPT-style model on 20 tokens staggering 120 million exaflops. This research...

Computed Tomography (CT) Image Reconstruction is an important technique used in a variety of domains, including medical imaging, electron microscopy, non-destructive testing and transportation security. Model-based Iterative (MBIR) using Coordinate Descent (ICD) CT algorithm that produces state-of-the-art results terms image quality. However, MBIR highly computationally intensive challenging to parallelize, has traditionally been viewed as impractical applications where reconstruction time...

Computed Tomographic (CT) image reconstruction is an important technique used in a wide range of applications. Among methods, Model-Based Iterative Reconstruction (MBIR) known to produce much higher quality CT images; however, the high computational requirements MBIR greatly restrict their application. Currently, speed primarily limited by irregular data access patterns, difficulty effective parallelization, and slow algorithmic convergence.

This paper considers the development of spatially adaptive smoothing splines for estimation a regression function with nonhomogeneous smoothness across domain. Two challenging issues arising in this context are evaluation equivalent kernel and determination local penalty. The penalty is design points order to accommodate behaviour function. We show that spline estimator approximately estimator, dependent. kernels traditional special case general solution. With aid Green's two-point boundary...

Computed Tomography (CT) serves as a key imaging technology that relies on computationally intensive filtering and back-projection algorithms for 3D image reconstruction. While conventional high-resolution reconstruction (> 2K3) solutions provide quick results, they typically treat an offline workload to be performed remotely large-scale HPC systems. The growing demand post-construction AI-driven analytics the need real-time adjustments call are feasible local computing resources, i.e....

Model-Based Image Reconstruction (MBIR) methods significantly enhance the quality of computed tomographic (CT) reconstructions relative to analytical techniques, but are limited by high computational cost. In this article, we propose a multi-agent consensus equilibrium (MACE) algorithm for distributing both computation and memory MBIR reconstruction across large number parallel nodes. MACE, each node stores only sparse subset views small portion system matrix, performs local sparse-view...

We present <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">FastConv</i> , a template-based code auto-generation open-source library that can automatically generate high-performance deep learning convolution kernels of arbitrary matrices/tensors shapes. FastConv is based on the Winograd algorithm, which reportedly highest performing algorithm for time-consuming layers convolutional neural networks. ARM CPUs cover wide range designs and...

Computed Tomography (CT) Image Reconstruction is an important technique used in a wide range of applications, ranging from explosive detection, medical imaging to scientific imaging. Among available reconstruction methods, Model Based Iterative (MBIR) produces higher quality images and allows for the use more general CT scanner geometries than possible with commonly methods. The high computational cost MBIR, however, often makes it impractical applications which would otherwise be ideal....

Iterative memory-bound solvers commonly occur in HPC codes. Typical GPU implementations have a loop on the host side that invokes kernel as much time/algorithm steps there are. The termination of each implicitly acts barrier required after advancing solution every time step. We propose an execution model for running iterative kernels: PERsistent KernelS (PERKS). In this model, is moved inside persistent kernel, and device-wide barriers are used synchronization. then reduce traffic to device...

Journal Article A class of grouped Brunk estimators and penalized spline for monotone regression Get access Xiao Wang, Wang Department Statistics, Purdue University, 250 N. University Street, West Lafayette, Indiana 47907, U.S.A.wangxiao@purdue.edu Search other works by this author on: Oxford Academic Google Scholar Jinglai Shen Mathematics Maryland Baltimore County, 1000 Hilltop Circle, Baltimore, 21250, U.S.A.shenj@umbc.edu Biometrika, Volume 97, Issue 3, September 2010, Pages 585–601,...

Computed tomography (CT) image reconstruction is a crucial technique for many imaging applications. Among various methods, Model-Based Iterative Reconstruction (MBIR) enables super-resolution with superior quality. MBIR, however, has high memory requirement that limits the achievable resolution, and parallelization MBIR suffers from limited scalability. In this paper, we propose Asynchronous Consensus (AC-MBIR) uses Equilibrium (CE) to provide algorithm small footprint, low communication...