PET is a popular medical imaging modality for various clinical applications, including diagnosis and image-guided radiation therapy. The low-dose (LDPET) at minimized dosage highly desirable in clinic since involves ionizing radiation, raises concerns about the risk of exposure. However, reduced dose radioactive tracers could impact image quality diagnosis. In this paper, supervised deep learning approach with generative adversarial network (GAN) cycle-consistency loss, Wasserstein distance...

The loop-mediated isothermal amplification (LAMP) assay indicates a potential and valuable means for genetically modified organism (GMO) detection especially its rapidity, simplicity, low cost. We developed evaluated the specificity sensitivity of LAMP method rapid (GM) maize T25. A set six specific primers was successfully designed to recognize distinct sequences on target gene, including pair inner primers, outer loop primers. optimum reaction temperature time were verified be 65°C 45 min,...

Super-resolution (SR) techniques have been employed to construct high-definition images from low-quality images. Various neural networks demonstrated excellent image-reconstruction quality in SR accelerators. However, deploying on edge devices is limited by resources and power consumption induced significant algorithm parameters, computation complexity, external memory accesses. This work explores the hardware co-design provide an end-to-end platform with a lightweight super-resolution...

Super-resolution (SR) networks have been gradually applied to embedded devices with good-quality image reconstruction. However, the hardware performance and power efficiency are limited by a large number of algorithm parameters, computation complexity, resources, obstructing development high-quality SR accelerator. This paper proposes an end-to-end platform lightweight super-resolution network (LSR) efficient, architecture HDSuper, perform algorithm-hardware co-design for For design, we...

Super-resolution (SR) based on deep learning has obtained superior performance in image reconstruction. Recently, various algorithm efforts have been committed to improving reconstruction quality and speed. However, the inference of SR contains huge amounts computation data access, leading low hardware implementation efficiency. For instance, up-sampling with deconvolution process requires considerable resources. In addition, sizes output feature maps several middle layers are...

Loop-mediated isothermal amplification (LAMP), a novel DNA technique, is widely used in molecular diagnosis although it very easy to be contaminated by the huge amount of its own amplicons. An improved visual LAMP method for event specific identification GM rice TT51-1 and endogenous reference gene (sucrose phosphate synthase, SPS) was established. Pre-loaded calcein manganous ion into reaction offered not only an way read results colour change from orange green but also closed- tube system...

Recently, the contrastive learning paradigm has achieved remarkable success in high-level tasks such as classification, detection, and segmentation. However, applied low-level tasks, like image restoration, is limited, its effectiveness uncertain. This raises a question: Why does not yield satisfactory results restoration? In this paper, we conduct in-depth analyses propose three guidelines to address above question. addition, inspired by style transfer based on learning, novel module for...

Along with the rapid evolution of deep neural networks, ever-increasing complexity imposes formidable computation intensity on hardware accelerator. In this paper, we propose a novel computing philosophy called "bit interleaving" and associate accelerator couple "Bitlet" Bitlet-X to maximally exploit bit-level sparsity. Apart from existing bit-serial/parallel accelerators, Bitlet leverages abundant "sparsity parallelism" in parameters enforce inference acceleration. is versatile by...

Deep neural networks have been deployed in various hardware accelerators, such as graph process units (GPUs), field-program gate arrays (FPGAs), and application specific integrated circuit (ASIC) chips. Normally, a huge amount of computation is required the inference process, creating significant logic resource overheads. In addition, frequent data accessions between off-chip memory accelerators create bottlenecks, leading to decline efficiency. Many solutions proposed reduce overhead...

With the development of algorithm and hardware, new computing paradigm changes from centralized intelligence on cloud to distributed edge computing. On hand, Artificial Intelligence (AI) has been applied a variety disciplines proved highly successful in vast class intelligent-edge applications cross many domains, such as wearable devices, biomedical monitoring systems, internet-of-things. hardware widely recognized promising solution meet requirements low latency, high scalability energy...

Computing-in-memory (CIM) architecture is a promising solution toward energy-efficient artificial intelligent (AI) processor. Practically, the AI processor with CIM engine induces series of issues including data updating and flexibility. For instance, in AI-oriented applications, weight stored must be reloaded due to huge gap between limited capacity growing parameter, which greatly reduces computation efficiency Moreover, natural parallelism leads mismatch various convolution kernel sizes...

Computing-in-memory (CIM) architecture has become a possible solution to designing an energy-efficient artificial intelligent processor. Various CIM demonstrators indicated the computing efficiency of macro and CIM-based processors. However, previous studies mainly focus on optimization low capacity without considering weight update strategy architecture. The intelligence (AI) processor with engine practically induces issues, including updating memory data supporting different operators. For...

Financial crisis is fatal to all companies. So, it very important establish a financial prediction system for each company resolve latent problems before they emerge. However, artificial neural network (ANN) offers an approach computation that different from conventional analytic methods. In this text, we construct BP model predict the of companies with samples Shanghai stock exchange and Shenzhen exchange. And proved appropriate.

Abstract Background The use of COOK balloon in the process induction labor is gradually promoted. This study was conducted to investigate safety and efficacy double dilation for pre-induction cervical maturation labor. Methods A total 343 pregnant women with full-term pregnancy Shanghai Putuo Maternity & Infant Health Hospital from January 1st September 30th 2016 were enrolled. Of all women, 166 had induction, which included a balloon, implementation artificial rupture membranes (AROM)...