Cluster Computing最新文献

In medical imaging, the classification and segmentation of lesions have always been significant topics in clinical research. Different categories of lesions require different treatment strategies, and accurate segmentation helps to assist in improving the effect of the clinical treatment. The Segment anything model (SAM) is an image segmentation model trained on a large-scale dataset with strong prompt segmentation capability, but it cannot be directly applied to the classification and segmentation tasks of medical images due to insufficient training on medical image data. In this paper, we propose a deep learning method for the classification and segmentation of lesions, called GMM-based segment anything model (G-SAM). Prompt-tuning is utilized in the model with the LoRA strategy, and the lesion feature extraction (GFE) module based on the Gaussian mixture model (GMM), is designed to effectively improve the effect of lesion classification and segmentation on the basis of the SAM. Notably, G-SAM exhibits greater sensitivity to early stage of the lesions, aiding in tumor detection and prevention, which holds important clinical value. G-SAM overcomes the limitation that SAM is not suitable for the medical image classification and segmentation tasks due to insufficient training data with minimal cost. Moreover, it enhances classification accuracy and segmentation precision compared to traditional Gaussian model-based methods. The effectiveness of G-SAM in classifying and segmenting lesions is validated on the LIDC dataset, demonstrating advantages over state-of-the-art (SOTA) methods. The study further validates the applicability of G-SAM on large publicly available datasets across three different image modalities, achieving superior performance.

Lupus nephritis (LN) is the most common symptom of systemic lupus erythematosus, emphasizing its importance in the field of medicine. The growing frequency of LN has increased the need for effective image segmentation algorithms. With the increasing prevalence of LN, the demand for efficient image segmentation techniques has grown. To enhance the efficiency of image segmentation of LN, many researchers employ a methodology that integrates multi-threshold image segmentation (MTIS) with metaheuristic algorithms (MAs). However, conventional MAs-based MTIS methods tend to converge towards local optima and have slow convergence rates, resulting in poor segmentation results within a limited iteration number. To address these challenges, this study proposes an advanced optimization algorithm termed Biogeography-based Learning Rime Optimization Algorithm (BLRIME) and integrates it with the MTIS approach for LN image segmentation. MTIS employs a non-local means 2D histogram to gather image information and uses 2D Renyi’s entropy as the fitness function. BLRIME builds upon the foundation of the RIME algorithm, incorporating two significant strategies. Firstly, the introduction of piecewise chaotic mapping (PCM) ameliorates the quality of the initial solution provided by the algorithm. Secondly, a stochastic biogeography-based learning strategy (SBLS) prevents the RIME algorithm from falling into the local optimum early. SBLS is proposed by this study based on the biogeography-based learning strategy. In order to assess the efficacy of the BLRIME, this paper devises a series of experiments to compare it with similar algorithms presented at IEEE CEC 2017. Experimental studies have been conducted to provide empirical evidence demonstrating the superior rates of convergence and precision achieved by BLRIME. Subsequently, the BLRIME-based MTIS algorithm is employed to segment the LN images compared to other peer algorithms. Furthermore, the peak signal-to-noise ratio, feature similarity index, and structural similarity index are utilized as evaluation metrics to assess the image segmentation outcomes. The experimental results prove that BLRIME demonstrates superior global search capabilities, resulting in remarkable outcomes in the segmentation of LN images.

This study analyzes the security implications of Proof-of-Work blockchains with respect to the stale block rate and the lack of a block verification process. The stale block rate is a crucial security metric that quantifies the proportion of rejected blocks in the blockchain network. The absence of a block verification process represents another critical security concern, as it permits the potential for invalid transactions within the network. In this article, we propose and implement a quantitative and analytical model to capture the primary operations of Proof-of-Work blockchains utilizing the Performance Evaluation Process Algebra. The proposed model can assist blockchain designers, architects, and analysts in achieving the ideal security level for blockchain systems by determining the proper network and consensus settings. We conduct extensive experiments to determine the sensitivity of security to four aspects: the number of active miners and their mining hash rates, the duration between blocks, the latency in block propagation, and the time required for block verification, all of which have been shown to influence the outcomes. We contribute to the findings of the existing research by conducting the first analysis of how the number of miners affects the frequency of stale block results, as well as how the delay in block propagation influences the incentives received by rational miners who choose to avoid the block verification process.

Given the problems with a centralized cloud and the emergence of ultra-low latency applications, and the needs of the Internet of Things (IoT), it has been found that novel methods are needed to support centralized cloud technology. Mobile edge computing is one of the solutions to mitigate these challenges. In this paper, we study task caching at Device to Device (D2D)-assisted network edge. In the proposed scheme, we predict the possibility of re-requesting tasks in the future using convolutional neural networks (CNN). Based on this predicted possibility, the number of last requests, and the number of cached versions of this task type in the neighbors, in addition to the characteristics of a task itself, including the required cache volume and processing resources, we rank the tasks using the proposed Multi-Criteria Task Ranking using Predicted requests (MCTRP) scheme and select the best replacement option in the cache of each Mobile User Equipment (MUE). The proposed scheme has proved to be of considerable benefit in terms of reducing delay and energy consumption and improving the utility of MUEs.

Classification as an essential part of Machine Learning and Data Mining has significant roles in engineering, medicine, agriculture, military, etc. With the evolution of data collection tools and the unceasing efforts of researchers, new datasets with huge dimensions are obtained so that each data sample has multiple labels. This kind of classification is called Multi-Class Classification (MLC) and demands new techniques to predict the set of labels for a data instance. To date, a variety of methods have been proposed to solve MLC problems. However, new high-dimensional datasets with challenging patterns are being developed, making it necessary for new research to be conducted to develop more efficient methods. This paper presents a novel framework named QLHA to solve MLC problems more efficiently. In the QLHA, the Principal Label Space Transformation (PLST) and Ridge Regression (RR) are recruited to predict the labels of data. Next, an effective objective function is introduced. Also, a hybrid metaheuristic algorithm called QGTOJS is provided to optimize objective value and enhance the predicted labels by selecting the most relevant features. In the QGTOJS, the Gorilla Troops Optimization (GTO) and Jellyfish Search algorithm (JS) are binarized and hybridized through a modified variant of the Q-learning algorithm. Besides, an adjusted Hill Climbing strategy is adopted to balance the exploration and exploitation and improve local optima departure. Likewise, a local search mechanism is provided to enhance searchability as much as possible. Eventually, the QLHA is applied to ten multi-label datasets and the obtained results are compared with heuristic and metaheuristic-based MLC methods numerically and visually. The experimental results disclosed the effectiveness of the contributions and superiority of the QLHA over competitors.

Oblivious transfer (OT) is a significant two party privacy preserving cryptographic primitive. OT involves a sender having several pieces of information and a receiver having a choice bit. The choice bit represents the piece of information that the receiver wants to obtain as an output of OT. At the end of the protocol, sender remains oblivious about the choice bit and receiver remains oblivious to the contents of the information that were not chosen. It has applications ranging from secure multi-party computation, privacy-preserving protocols to cryptographic protocols for secure communication. Most of the classical OT protocols are based on number theoretic assumptions which are not quantum secure and existing quantum OT protocols are not so efficient and practical. Herein, we present the design and analysis of a simple yet efficient quantum OT protocol, namely qOT. qOT is designed by using the asymmetric key distribution proposed by Gao et al. (Opt Express 20(16):17411–17420, 2012) as a building block. The designed qOT requires only single photons as a source of a quantum state, and the measurements of the states are computed using single particle projective measurement. These make qOT efficient and practical. Our proposed design is secure against quantum attacks. Moreover, qOT also provides long-term security.