IEEE Access最新文献_第6页

Assessing the Value of Heterogeneous Elasticities for Incentive-Based Residential Demand Response 基于激励的住宅需求响应的异质性弹性价值评估

IF 3.4 3区计算机科学

IEEE Access Pub Date : 2025-07-22 DOI: 10.1109/ACCESS.2025.3591723

Bahareh Kargar;Elson Cibaku;Sangwoo Park

{"title":"Assessing the Value of Heterogeneous Elasticities for Incentive-Based Residential Demand Response","authors":"Bahareh Kargar;Elson Cibaku;Sangwoo Park","doi":"10.1109/ACCESS.2025.3591723","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3591723","url":null,"abstract":"Incentive-based demand response (IBDR) programs play a crucial role in enhancing grid stability and reducing peak loads in modern power systems. However, existing IBDR programs often rely on aggregate demand models, overlooking the impact of heterogeneous consumer behaviors and appliance-specific demand elasticities. This study assesses the value of incorporating heterogeneous elasticity values in IBDR programs by developing three optimization models with increasing levels of granularity: 1) an aggregate elasticity model, 2) an appliance-specific elasticity model, and 3) a customer and appliance-specific elasticity model. Furthermore, this study incorporates transmission line losses into the models, providing a realistic assessment of distribution system efficiency. Comparative analysis using realistic residential electricity consumption data reveals that integrating appliance-specific elasticity significantly improves economic efficiency, while adding customer-specific granularity yields marginal additional benefits. Comparative analysis using realistic residential electricity consumption data reveals that integrating appliance-specific elasticity improves economic efficiency by 6.29%, while customer-specific granularity yields only a marginal additional benefit of 0.92%. These findings offer valuable insights for load-serving entities (LSEs) and policymakers in designing more efficient IBDR programs.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"129900-129910"},"PeriodicalIF":3.4,"publicationDate":"2025-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11088113","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144716216","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Enhanced Spectral Efficiency in RIS-Assisted MIMO Systems Through Joint Precoding and RIS Configuration 通过联合预编码和RIS配置提高RIS辅助MIMO系统的频谱效率

IF 3.4 3区计算机科学

IEEE Access Pub Date : 2025-07-22 DOI: 10.1109/ACCESS.2025.3590964

Mohammed Adil Abbas;Aqiel N. Almamori;Ahmed Jumaa Lafta;Asaad H. Sahar

{"title":"Enhanced Spectral Efficiency in RIS-Assisted MIMO Systems Through Joint Precoding and RIS Configuration","authors":"Mohammed Adil Abbas;Aqiel N. Almamori;Ahmed Jumaa Lafta;Asaad H. Sahar","doi":"10.1109/ACCESS.2025.3590964","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3590964","url":null,"abstract":"Reconfigurable Intelligent Surfaces (RIS) offer a promising solution to enhance wireless communication performance, particularly for 6G networks. This paper proposes a unified joint optimization framework to enhance spectral efficiency in reconfigurable intelligent surface (RIS)-assisted multiple-input multiple-output (MIMO) systems. Unlike conventional approaches that optimize RIS phase shifts and transmitter precoding separately, our method jointly optimizes both using an iterative strategy. Water-filling is used for power allocation across channel eigenmodes, while manifold optimization ensures efficient phase shift updates under unit-modulus constraints. Extensive simulations under diverse channel conditions reveal a consistent spectral efficiency improvement of up to 39.38%, outperforming across RIS sizes and transmit powers. These results highlight the contribution of combining RIS configuration and transmitter precoding into a coordinated optimization loop, guided by channel eigenmode alignment and practical implementation constraints. The enhanced performance stems from the algorithm’s ability to dynamically coordinate power allocation with channel eigenmode alignment, making it a viable solution for next-generation wireless systems requiring high spectral efficiency.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"130159-130167"},"PeriodicalIF":3.4,"publicationDate":"2025-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11086580","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144716322","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Soft Error in Saddle Fin-Based DRAM at Cryogenic Temperature 低温下鞍片型DRAM的软误差研究

IF 3.4 3区计算机科学

IEEE Access Pub Date : 2025-07-22 DOI: 10.1109/ACCESS.2025.3591488

Minsang Ryu;Minki Suh;Jonghyeon Ha;Minji Bang;Dabok Lee;Hojoon Lee;Hyunchul Sagong;Dong-Seok Kim;Jungsik Kim

{"title":"Soft Error in Saddle Fin-Based DRAM at Cryogenic Temperature","authors":"Minsang Ryu;Minki Suh;Jonghyeon Ha;Minji Bang;Dabok Lee;Hojoon Lee;Hyunchul Sagong;Dong-Seok Kim;Jungsik Kim","doi":"10.1109/ACCESS.2025.3591488","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3591488","url":null,"abstract":"This study examines the impact of soft error by heavy ions on saddle fin-based dynamic random access memory (DRAM). The investigation is conducted using technology computer-aided design (TCAD) simulation at different temperatures ranging from 77 to 300 K. At 300 K, charge sharing is greater compared to 77 K due to the increased prominence of the bipolar amplification effect. The decrease in storage node potential (<inline-formula> <tex-math>$V_{SN}$ </tex-math></inline-formula>) caused by charge sharing varies by up to 1.31% between 77 and 300 K. Nevertheless, if the linear energy transfer (LET) of the ion is below 1 MeV<inline-formula> <tex-math>$cdot $ </tex-math></inline-formula>cm2/mg, temperature increase does not result in enhanced charge sharing. This is because there is an insufficient generation of electron-hole pairs (EHPs) to trigger a bipolar amplification effect. On the other hand, the amount collected charge is greater at 77 compared to 300 K because the mobility of the carriers increased as the temperature decreased. The variation in <inline-formula> <tex-math>$V_{SN}$ </tex-math></inline-formula> due to the collected charge is as high as 13.19% between 77 and 300 K. When comparing the reduction in <inline-formula> <tex-math>$V_{SN}$ </tex-math></inline-formula> caused by collected charge and charge sharing, it is seen that the influence of collected charge is more pronounced at 77 and 300 K. TCAD simulations are used to investigate strategies for mitigating the heavy ion effect. Enhancing the bit-line junction can reduce the impact of heavy ions on the saddle fin-based DRAM. As a result, several EHPs generated by heavy ions can be moved towards the junction of the bit-line.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"130603-130609"},"PeriodicalIF":3.4,"publicationDate":"2025-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11088077","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144716328","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Advances in Virtual Power Plant Operations: A Review of Optimization Models 虚拟电厂运行研究进展：优化模型综述

IF 3.6 3区计算机科学

IEEE Access Pub Date : 2025-07-22 DOI: 10.1109/ACCESS.2025.3591697

Fatemeh Marzbani;Ahmed H. Osman;Mohamed S. Hassan

{"title":"Advances in Virtual Power Plant Operations: A Review of Optimization Models","authors":"Fatemeh Marzbani;Ahmed H. Osman;Mohamed S. Hassan","doi":"10.1109/ACCESS.2025.3591697","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3591697","url":null,"abstract":"In recent decades, the global shift toward sustainable energy solutions has accelerated, prompting nations to integrate renewable energy sources (RES) into their electricity grids and adhere to international environmental protocols. This transition has catalyzed the evolution from traditional power systems to smart grids, which form the backbone of contemporary energy management systems. However, smart grids often show limited adaptability to the large-scale integration of RES, underscoring a critical need for more dynamic solutions. Against this backdrop, Virtual Power Plants (VPPs) have emerged as a pivotal innovation, enhancing grid management through their robust handling of RES fluctuations and sophisticated distributed control systems. VPPs not only increase the flexibility, reliability, scalability, and performance of power systems but also improve asset utilization, bolster system resilience, and foster more effective market interactions. This surge in VPP deployment has sparked extensive research focused on optimizing VPP operations and their engagement with electricity markets. Given the rapid advancements in VPP technology and market integration, this review is critical for consolidating existing knowledge, guiding effective implementation strategies, and identifying emerging trends and challenges within the field. This review presents a comprehensive overview of prior research on VPP optimization, delineating the key methodologies and outcomes that have shaped current practices. Specifically, this paper discusses the fundamental concepts of VPPs, provides an overview of their integration into electricity markets, and examines the various optimization formulations and methodologies that have been proposed in the literature for enhancing VPP operational efficiency.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"131525-131548"},"PeriodicalIF":3.6,"publicationDate":"2025-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11088101","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144725306","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A Comparative Study for Localization of Forgery Regions in Images 图像伪造区域定位的比较研究

IF 3.4 3区计算机科学

IEEE Access Pub Date : 2025-07-22 DOI: 10.1109/ACCESS.2025.3591571

Mustafa Özden;Canberk Şahin

{"title":"A Comparative Study for Localization of Forgery Regions in Images","authors":"Mustafa Özden;Canberk Şahin","doi":"10.1109/ACCESS.2025.3591571","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3591571","url":null,"abstract":"As computer technologies and image processing software have advanced, it has become progressively easier to produce simple fake or forged images by altering digital images without leaving any discernible trace. There is a significant need to detect manipulated regions in images in crucial fields such as politics, law, and forensic medicine. In this study, we propose a method that combines the traditional techniques, such as Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT), with the advantages of deep learning methods to detect manipulated regions in forged images. The proposed method involves designing an architecture where DWT and DCT are used in parallel with DenseNet based Convolutional Neural Network (CNN). To evaluate the effectiveness of this method, we implemented three alternative approaches: one that uses only DCT and CNN, another that uses only DWT and CNN, and a third that employs only CNN without either transformation. In total, four different methods were tested on eight datasets, and their performance was compared using metrics such as accuracy, precision, recall, dice similarity coefficient, and F1 score. The results from these comparisons clearly indicate the effectiveness and high classification accuracy of the proposed method. By leveraging the combined strengths of traditional image processing techniques and advanced deep learning algorithms, the proposed method demonstrates superior capability in detecting manipulated regions in forged images, thus offering a robust solution for applications in forensic field.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"130701-130718"},"PeriodicalIF":3.4,"publicationDate":"2025-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11088100","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144716170","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Autoencoder-Augmented Graph Neural Networks for Accurate and Scalable Structure Recognition in Analog/Mixed-Signal Schematics 用于精确和可扩展的模拟/混合信号原理图结构识别的自编码器增强图神经网络

IF 3.4 3区计算机科学

IEEE Access Pub Date : 2025-07-22 DOI: 10.1109/ACCESS.2025.3591720

Mohamed Salem;Witesyavwirwa Vianney Kambale;Ali Deeb;Sergii Tkachov;Anjeza Karaj;Joachim Pichler;Manuel Ludwig Lexer;Kyandoghere Kyamakya

{"title":"Autoencoder-Augmented Graph Neural Networks for Accurate and Scalable Structure Recognition in Analog/Mixed-Signal Schematics","authors":"Mohamed Salem;Witesyavwirwa Vianney Kambale;Ali Deeb;Sergii Tkachov;Anjeza Karaj;Joachim Pichler;Manuel Ludwig Lexer;Kyandoghere Kyamakya","doi":"10.1109/ACCESS.2025.3591720","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3591720","url":null,"abstract":"The increasing complexity of Analog/Mixed-Signal (AMS) schematics has been posing significant challenges in structure recognition, particularly in the intellectual property (IP) industry, where data scarcity and confidentiality constraints limit model training. In this work, a novel framework has been proposed that combines the generative augmentation capabilities of convolutional Autoencoders with the structural analysis power of Graph Convolutional Networks (GCNs). Realistic schematic variants have been synthesized from limited proprietary data to enhance model generalization, while the GCN has been used to capture topological features critical to substructure recognition. The method has been validated on a curated AMS dataset, where it surpassed a GCN-only baseline by reducing reconstruction error and achieving a balanced classification accuracy of 96.7%, thereby exceeding the long-standing 95% accuracy threshold. Inference latency was measured at 5–10ms per schematic on standard GPU hardware, confirming its applicability to interactive industrial Electronic Design Automation (EDA) workflows. These results highlight the potential of the Autoencoder–GCN pipeline as a scalable and reliable solution for AMS structure recognition under real-world constraints.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"129721-129740"},"PeriodicalIF":3.4,"publicationDate":"2025-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11088082","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144716273","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Adaptive Federated Learning With Reinforcement Learning-Based Client Selection for Heterogeneous Environments 异构环境下基于强化学习的自适应联邦学习客户端选择

IF 3.6 3区计算机科学

IEEE Access Pub Date : 2025-07-22 DOI: 10.1109/ACCESS.2025.3591699

Shamim Ahmed;M. Shamim Kaiser;Sudipto Chaki;A. B. M. Shawkat Ali

{"title":"Adaptive Federated Learning With Reinforcement Learning-Based Client Selection for Heterogeneous Environments","authors":"Shamim Ahmed;M. Shamim Kaiser;Sudipto Chaki;A. B. M. Shawkat Ali","doi":"10.1109/ACCESS.2025.3591699","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3591699","url":null,"abstract":"This study introduces an Adaptive Federated Learning (AFL) framework designed to address the challenges of data heterogeneity, resource imbalance, and communication constraints in decentralized learning environments. The framework integrates reinforcement learning (RL) based client selection using both Tabular Q-Learning and Deep Q-Network (DQN) strategies to dynamically identify clients that most positively impact global model performance. A multi-objective reward function, combining model accuracy and execution time, guides the RL agent toward performance- and efficiency-aware client selection. For local model training, Random Forest (RF) classifiers are employed to ensure robustness to noise, class imbalance, and limited computational resources, particularly in privacy-sensitive healthcare settings. The AFL framework is evaluated on two real-world healthcare datasets BRFSS2015 and Diabetes Prediction, and extended to benchmark FL datasets (CIFAR-10 and FEMNIST) to assess scalability and generalization. Experimental results demonstrate that the DQN-based AFL achieves superior global accuracy (up to 91.3%) compared to Tabular Q-Learning and baseline methods such as FedAvg, while also reducing execution time by up to 15%. Client-level accuracy remains stable across rounds, with reward progression confirming effective RL policy convergence. These findings underscore the AFL framework’s capability to adaptively balance performance and efficiency, offering a practical and scalable solution for federated learning in heterogeneous, resource-constrained environments.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"131671-131695"},"PeriodicalIF":3.6,"publicationDate":"2025-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11088109","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144725252","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Explainable Graph Neural Networks for Power Grid Fault Detection 用于电网故障检测的可解释图神经网络

IF 3.6 3区计算机科学

IEEE Access Pub Date : 2025-07-22 DOI: 10.1109/ACCESS.2025.3591604

Richard Bosso;Corey Chang;Mahdi Zarif;Yufei Tang

{"title":"Explainable Graph Neural Networks for Power Grid Fault Detection","authors":"Richard Bosso;Corey Chang;Mahdi Zarif;Yufei Tang","doi":"10.1109/ACCESS.2025.3591604","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3591604","url":null,"abstract":"This paper proposes the application of explanation methods to enhance the interpretability of graph neural network (GNN) models in fault location for power grids. GNN models have exhibited remarkable precision in utilizing phasor data from various locations around the grid and integrating the system’s topology, an advantage rarely harnessed by alternative machine learning techniques. This capability makes GNNs highly effective in identifying fault occurrences in power grids. Despite their greater performance, these models can encounter criticism for their “black box” nature, which conceals the reasoning behind their predictions. Lack of transparency significantly hinders power utility operations, as interpretability is crucial to building trust, accountability, and actionable insights. This research presents a comprehensive framework that systematically evaluates state-of-the-art explanation strategies, representing the first use of such a framework for Graph Neural Network models for defect location detection. By assessing the strengths and weaknesses of different explanatory methods, it identifies and recommends the most effective strategies for clarifying the decision-making processes of GNN models. These recommendations aim to improve the transparency of fault predictions, allowing utility providers to better understand and trust the models’ output. The proposed framework not only enhances the practical usability of GNN-based systems but also contributes to advancing their adoption in critical power grid applications.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"129520-129533"},"PeriodicalIF":3.6,"publicationDate":"2025-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11088107","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144725307","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

MITSGRN: A Novel Computational Framework for Reconstructing Sleep Rhythm Gene Regulatory Networks Based on Mutual Information and Time-Series Big Data MITSGRN：基于互信息和时间序列大数据重构睡眠节律基因调控网络的新计算框架

IF 3.4 3区计算机科学

IEEE Access Pub Date : 2025-07-22 DOI: 10.1109/ACCESS.2025.3591304

Zhenyu Liu;Jiangqian Zuo;Qian Cao;Zheng Lu;Tao Li

{"title":"MITSGRN: A Novel Computational Framework for Reconstructing Sleep Rhythm Gene Regulatory Networks Based on Mutual Information and Time-Series Big Data","authors":"Zhenyu Liu;Jiangqian Zuo;Qian Cao;Zheng Lu;Tao Li","doi":"10.1109/ACCESS.2025.3591304","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3591304","url":null,"abstract":"Disruptions in sleep rhythms have emerged as a global health concern, posing serious risks to the physical and mental well-being of modern populations. Elucidating the molecular regulatory mechanisms underlying the periodic nature of sleep rhythms remains a critical scientific challenge. In this study, we propose an innovative computational framework for gene regulatory network (GRN) reconstruction based on mutual information and large-scale time-series data. The proposed framework leverages the temporal characteristics of gene expression profiles associated with sleep rhythms, and integrates k-means clustering, mutual information, and Pearson lag correlation analysis in a synergistic manner to support GRN reconstruction. We systematically evaluate the performance of our method using BEELINE open-source datasets of varying scales, with precision, recall, and cross-validation accuracy as evaluation metrics. Experimental results demonstrate that our approach significantly outperforms existing methods such as dynGENIE3 and transfer entropy in terms of both accuracy and generalization capability. Furthermore, we successfully applied the proposed framework to reconstruct the GRN governing sleep rhythms in rats. The resulting network exhibits topological features and identifies key regulatory components that are highly consistent with previously published findings. Our results highlight the advantages of mutual information-based GRN reconstruction in deciphering complex biological rhythm regulatory systems. This method not only provides a novel perspective for investigating the gene regulatory mechanisms underlying sleep rhythms, but also establishes a solid methodological foundation for exploring the pathogenesis of sleep-related disorders and advancing the development of targeted therapies.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"130088-130097"},"PeriodicalIF":3.4,"publicationDate":"2025-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11087579","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144716179","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Dual-Stream Contrastive Learning for Medical Visual Representations Using Synthetic Images Generated by Latent Diffusion Model 基于潜在扩散模型合成图像的医学视觉表征双流对比学习

IF 3.6 3区计算机科学

IEEE Access Pub Date : 2025-07-22 DOI: 10.1109/ACCESS.2025.3591544

Weitao Ye;Longfu Zhang;Xiaoben Jiang;Dawei Yang;Yu Zhu

{"title":"Dual-Stream Contrastive Learning for Medical Visual Representations Using Synthetic Images Generated by Latent Diffusion Model","authors":"Weitao Ye;Longfu Zhang;Xiaoben Jiang;Dawei Yang;Yu Zhu","doi":"10.1109/ACCESS.2025.3591544","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3591544","url":null,"abstract":"Deep learning-based medical image processing methods can enhance diagnostic accuracy while significantly accelerating clinical decision workflows. However, in order to learn better visual representations, such approaches usually need substantial amount of expert-annotated data, which are highly costly. To address this issue, we propose a novel approach called Dual-Stream Contrastive Learning with Cross-Scale Token Projection (DCL-CsTP), which aims to enhance visual representations and transferable initializations. Specifically, a latent diffusion model (LDM) is leveraged to generate high-quality synthetic medical images in order to expand the dataset. Then we utilize the proposed dual-stream architecture that consists of a global semantic relations stream and a local detail relations stream to learn discriminative medical image representations from the dataset. Furthermore, a cross-scale token projection is designed to enable the model to capture various scales of focus in medical images. Comprehensive experiments are performed on two downstream tasks: medical image classification and segmentation. For multi-classification of pneumonia, our DCL-CsTP method achieves 95.90% accuracy. For lesions segmentation, our DCL-CsTP method attains 89.73% dice coefficient on the International Skin Imaging Collaboration 2018 (ISIC 2018) dataset and 82.50% dice coefficient on the Kvasir-SEG dataset. The performance superiority of the model pre-trained by DCL-CsTP is conclusively demonstrated through the above experiments on various dataset, which shows that DCL-CsTP can enhance diagnostic precision and alleviate radiologists’ image screening burdens.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"129648-129658"},"PeriodicalIF":3.6,"publicationDate":"2025-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11088093","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144725256","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0