Pub Date : 2024-04-12DOI: 10.1109/TETCI.2024.3381512
Zhitong Ma;Jinghui Zhong;Wei-Li Liu;Jun Zhang
Evolutionary multitasking(EMT), which conducts evolutionary research on multiple tasks simultaneously, is an emerging research topic in the computation intelligence community. It aims to enhance the convergence characteristics by simultaneously conducting evolutionary research on multiple tasks, thereby facilitating knowledge transfer among tasks and achieving exceptional performance in solution quality. However, most of the existing EMT algorithms still suffer from the high computational burden especially when the number of tasks is large. To address this issue, this paper proposes a GPU-based multitasking evolutionary framework, which is able to handle thousands of tasks that arrive asynchronous in a short time. Besides, a concurrent multi-island mechanism is proposed to enable the parallel EMT algorithm to efficiently solve high-dimensional problems. Experimental results on eight problems with differing characteristics have demonstrated that the proposed framework is effective in solving high-dimensional problems and can significantly reduce the search time.
{"title":"Accelerating Evolutionary Multitasking Optimization With a Generalized GPU-Based Framework","authors":"Zhitong Ma;Jinghui Zhong;Wei-Li Liu;Jun Zhang","doi":"10.1109/TETCI.2024.3381512","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3381512","url":null,"abstract":"Evolutionary multitasking(EMT), which conducts evolutionary research on multiple tasks simultaneously, is an emerging research topic in the computation intelligence community. It aims to enhance the convergence characteristics by simultaneously conducting evolutionary research on multiple tasks, thereby facilitating knowledge transfer among tasks and achieving exceptional performance in solution quality. However, most of the existing EMT algorithms still suffer from the high computational burden especially when the number of tasks is large. To address this issue, this paper proposes a GPU-based multitasking evolutionary framework, which is able to handle thousands of tasks that arrive asynchronous in a short time. Besides, a concurrent multi-island mechanism is proposed to enable the parallel EMT algorithm to efficiently solve high-dimensional problems. Experimental results on eight problems with differing characteristics have demonstrated that the proposed framework is effective in solving high-dimensional problems and can significantly reduce the search time.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"8 6","pages":"3995-4010"},"PeriodicalIF":5.3,"publicationDate":"2024-04-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142691730","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-04-11DOI: 10.1109/TETCI.2024.3382233
Luqiao Li;Zhihua Chen;Lei Dai;Ran Li;Bin Sheng
High-quality clear images are the basis for advanced vision tasks such as target detection and semantic segmentation. This paper proposes an image dehazing algorithm named mixed attention-based multi-scale feature calibration network, aiming at solving the problem of uneven haze distribution in low-quality fuzzy images acquired in foggy environments, which is difficult to remove effectively. Our algorithm adopts a U-shaped structure to extract multi-scale features and deep semantic information. In the encoding module, a mixed attention module is designed to assign different weights to each position in the feature map, focusing on the important information and regions where haze is difficult to be removed in the image. In the decoding module, a self-calibration recovery module is designed to fully integrate different levels of features, calibrate feature information, and restore spatial texture details. Finally, the multi-scale feature information is aggregated by the reconstruction module and accurately mapped into the solution space to obtain a clear image after haze removal. Extensive experiments show that our algorithm outperforms state-of-the-art image dehazing algorithms in various synthetic datasets and real hazy scenes in terms of qualitative and quantitative comparisons, and can effectively remove haze in different scenes and recover images with high quality.
高质量的清晰图像是目标检测和语义分割等高级视觉任务的基础。本文提出了一种名为 "基于混合注意力的多尺度特征校准网络 "的图像去污算法，旨在解决雾霾环境下获取的低质量模糊图像中雾霾分布不均匀、难以有效去除的问题。我们的算法采用 U 型结构提取多尺度特征和深层语义信息。在编码模块中，设计了一个混合注意力模块，为特征图中的每个位置分配不同的权重，重点关注图像中的重要信息和难以去除雾霾的区域。在解码模块中，设计了一个自校准恢复模块，以充分整合不同层次的特征，校准特征信息，恢复空间纹理细节。最后，多尺度特征信息由重构模块汇总，并精确映射到解算空间，从而获得去除雾霾后的清晰图像。大量实验表明，在各种合成数据集和真实雾霾场景中，我们的算法在定性和定量比较方面都优于最先进的图像去雾霾算法，能有效去除不同场景中的雾霾，恢复出高质量的图像。
{"title":"MA-MFCNet: Mixed Attention-Based Multi-Scale Feature Calibration Network for Image Dehazing","authors":"Luqiao Li;Zhihua Chen;Lei Dai;Ran Li;Bin Sheng","doi":"10.1109/TETCI.2024.3382233","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3382233","url":null,"abstract":"High-quality clear images are the basis for advanced vision tasks such as target detection and semantic segmentation. This paper proposes an image dehazing algorithm named mixed attention-based multi-scale feature calibration network, aiming at solving the problem of uneven haze distribution in low-quality fuzzy images acquired in foggy environments, which is difficult to remove effectively. Our algorithm adopts a U-shaped structure to extract multi-scale features and deep semantic information. In the encoding module, a mixed attention module is designed to assign different weights to each position in the feature map, focusing on the important information and regions where haze is difficult to be removed in the image. In the decoding module, a self-calibration recovery module is designed to fully integrate different levels of features, calibrate feature information, and restore spatial texture details. Finally, the multi-scale feature information is aggregated by the reconstruction module and accurately mapped into the solution space to obtain a clear image after haze removal. Extensive experiments show that our algorithm outperforms state-of-the-art image dehazing algorithms in various synthetic datasets and real hazy scenes in terms of qualitative and quantitative comparisons, and can effectively remove haze in different scenes and recover images with high quality.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"8 5","pages":"3408-3421"},"PeriodicalIF":5.3,"publicationDate":"2024-04-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142368262","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-04-10DOI: 10.1109/TETCI.2024.3382232
Shaofu Yang;Yang Shen;Jinde Cao;Tingwen Huang
In this paper, we consider the problem of distributed optimization over time-varying directed graphs, where each agent maintains a private objective function and the goal of all agents is to cooperatively minimize the sum of their objects. By combining H eavy- B all method with B arzilai- B orwein step sizes, a novel discrete-time accelerated distributed algorithm termed as HBBB is proposed. Compared with existing distributed algorithms over digraphs, HBBB exploits the merit of both momentum and adaptive step sizes for acceleration but requiring very few additional computational costs. It is proved that the algorithm converges to the exact optimal solution at a geometric rate as long as a scale factor for step sizes and the momentum coefficient do not exceed certain bounds. Finally, numerical experiments are performed to show the effectiveness of HBBB as well as its performance, which is comparable to and sometimes even better than existing distributed optimization algorithms over general directed graphs.
{"title":"Distributed Heavy-Ball Over Time-Varying Digraphs With Barzilai-Borwein Step Sizes","authors":"Shaofu Yang;Yang Shen;Jinde Cao;Tingwen Huang","doi":"10.1109/TETCI.2024.3382232","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3382232","url":null,"abstract":"In this paper, we consider the problem of distributed optimization over time-varying directed graphs, where each agent maintains a private objective function and the goal of all agents is to cooperatively minimize the sum of their objects. By combining \u0000<underline>H</u>\u0000eavy-\u0000<underline>B</u>\u0000all method with \u0000<underline>B</u>\u0000arzilai-\u0000<underline>B</u>\u0000orwein step sizes, a novel discrete-time accelerated distributed algorithm termed as HBBB is proposed. Compared with existing distributed algorithms over digraphs, HBBB exploits the merit of both momentum and adaptive step sizes for acceleration but requiring very few additional computational costs. It is proved that the algorithm converges to the exact optimal solution at a geometric rate as long as a scale factor for step sizes and the momentum coefficient do not exceed certain bounds. Finally, numerical experiments are performed to show the effectiveness of HBBB as well as its performance, which is comparable to and sometimes even better than existing distributed optimization algorithms over general directed graphs.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"8 6","pages":"4011-4021"},"PeriodicalIF":5.3,"publicationDate":"2024-04-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142691758","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-04-10DOI: 10.1109/TETCI.2024.3372420
Heli Sun;Ruirui Xue;Tingting Hu;Tengfei Pan;Liang He;Yuan Rao;Zhi Wang;Yingxue Wang;Yuan Chen;Hui He
Citywide crowd flow prediction is an important problem for traffic control, risk assessment, and public safety, especially in critical areas. However, the large scale of the city and the interactions between multiple regions make this problem more challenging. Furthermore, it is impacted by temporal closeness, period, and trend features. Besides, geographic information and meta-features, such as periods of a day and days of a week also affect spatio-temporal correlation. Simultaneously, the influence between different regions will change over time, which is called dynamic correlation. We concentrate on how to concurrently model the important features and dynamic spatial correlation to increase prediction accuracy and simplify the problem. To forecast the crowd flow in critical areas, we propose a two-step framework. First, the grid density peak clustering algorithm is used to set the temporal attenuation factor, which selects the critical areas. Then, the effects of geographic information on spatio-temporal correlation are modeled by graph embedding and the effects of different temporal features are represented by graph convolutional neural networks. In addition, we use the multi-attention mechanism to capture the dynamic spatio-temporal correlation. On two real datasets, experimental results show that our model can balance time complexity and prediction accuracy well. It is 20% better in accuracy than other baselines, and the prediction speed is better than most models.
{"title":"Predicting Citywide Crowd Flows in Critical Areas Based on Dynamic Spatio-Temporal Network","authors":"Heli Sun;Ruirui Xue;Tingting Hu;Tengfei Pan;Liang He;Yuan Rao;Zhi Wang;Yingxue Wang;Yuan Chen;Hui He","doi":"10.1109/TETCI.2024.3372420","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3372420","url":null,"abstract":"Citywide crowd flow prediction is an important problem for traffic control, risk assessment, and public safety, especially in critical areas. However, the large scale of the city and the interactions between multiple regions make this problem more challenging. Furthermore, it is impacted by temporal closeness, period, and trend features. Besides, geographic information and meta-features, such as periods of a day and days of a week also affect spatio-temporal correlation. Simultaneously, the influence between different regions will change over time, which is called dynamic correlation. We concentrate on how to concurrently model the important features and dynamic spatial correlation to increase prediction accuracy and simplify the problem. To forecast the crowd flow in critical areas, we propose a two-step framework. First, the grid density peak clustering algorithm is used to set the temporal attenuation factor, which selects the critical areas. Then, the effects of geographic information on spatio-temporal correlation are modeled by graph embedding and the effects of different temporal features are represented by graph convolutional neural networks. In addition, we use the multi-attention mechanism to capture the dynamic spatio-temporal correlation. On two real datasets, experimental results show that our model can balance time complexity and prediction accuracy well. It is 20% better in accuracy than other baselines, and the prediction speed is better than most models.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"8 5","pages":"3703-3715"},"PeriodicalIF":5.3,"publicationDate":"2024-04-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142377137","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-04-08DOI: 10.1109/TETCI.2024.3382230
Feilong Cao;Xing Ye;Hailiang Ye
Graph neural networks are easily deceived by adversarial attacks that intentionally modify the graph structure. Particularly, homophilous edges connecting similar nodes can be maliciously deleted when adversarial edges are inserted into the graph. Graph structure learning (GSL) reconstructs an optimal graph structure and corresponding representation and has recently received considerable attention in adversarial attacks. However, constrained by a single topology view of the poisoned graph and few labels, most GSL techniques are difficult to effectively learn robust representations that sufficiently carry precise structure information and similar node information. Therefore, this paper develops a robust multi-view graph contrastive learning (RM-GCL) framework to defend against adversarial attacks. It exploits additional structural information and contrastive supervision signals from the data to guide graph structure optimization. In particular, an adaptive graph-augmented contrastive learning (AGCL) module is devised to obtain reliable representations. Besides, a node-level attention mechanism is incorporated to fuse these representations adaptively acquired from AGCL and then complete node classification tasks. Experiments on multiple datasets manifest that RM-GCL exceeds the state-of-the-art approaches and successfully defends against various attacks.
{"title":"A Multi-View Graph Contrastive Learning Framework for Defending Against Adversarial Attacks","authors":"Feilong Cao;Xing Ye;Hailiang Ye","doi":"10.1109/TETCI.2024.3382230","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3382230","url":null,"abstract":"Graph neural networks are easily deceived by adversarial attacks that intentionally modify the graph structure. Particularly, homophilous edges connecting similar nodes can be maliciously deleted when adversarial edges are inserted into the graph. Graph structure learning (GSL) reconstructs an optimal graph structure and corresponding representation and has recently received considerable attention in adversarial attacks. However, constrained by a single topology view of the poisoned graph and few labels, most GSL techniques are difficult to effectively learn robust representations that sufficiently carry precise structure information and similar node information. Therefore, this paper develops a robust multi-view graph contrastive learning (RM-GCL) framework to defend against adversarial attacks. It exploits additional structural information and contrastive supervision signals from the data to guide graph structure optimization. In particular, an adaptive graph-augmented contrastive learning (AGCL) module is devised to obtain reliable representations. Besides, a node-level attention mechanism is incorporated to fuse these representations adaptively acquired from AGCL and then complete node classification tasks. Experiments on multiple datasets manifest that RM-GCL exceeds the state-of-the-art approaches and successfully defends against various attacks.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"8 6","pages":"4022-4032"},"PeriodicalIF":5.3,"publicationDate":"2024-04-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142691736","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-04-04DOI: 10.1109/TETCI.2024.3379240
Jie Xu;Zihan Wu;Cong Wang;Xiaohua Jia
Machine learning models may inadvertently memorize sensitive, unauthorized, or malicious data, posing risks of privacy breaches, security vulnerabilities, and performance degradation. To address these issues, machine unlearning has emerged as a critical technique to selectively remove specific training data points' influence on trained models. This paper provides a comprehensive taxonomy and analysis of the solutions in machine unlearning. We categorize existing solutions into exact unlearning approaches that remove data influence thoroughly and approximate unlearning approaches that efficiently minimize data influence. By comprehensively reviewing solutions, we identify and discuss their strengths and limitations. Furthermore, we propose future directions to advance machine unlearning and establish it as an essential capability for trustworthy and adaptive machine learning models. This paper provides researchers with a roadmap of open problems, encouraging impactful contributions to address real-world needs for selective data removal.
{"title":"Machine Unlearning: Solutions and Challenges","authors":"Jie Xu;Zihan Wu;Cong Wang;Xiaohua Jia","doi":"10.1109/TETCI.2024.3379240","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3379240","url":null,"abstract":"Machine learning models may inadvertently memorize sensitive, unauthorized, or malicious data, posing risks of privacy breaches, security vulnerabilities, and performance degradation. To address these issues, machine unlearning has emerged as a critical technique to selectively remove specific training data points' influence on trained models. This paper provides a comprehensive taxonomy and analysis of the solutions in machine unlearning. We categorize existing solutions into exact unlearning approaches that remove data influence thoroughly and approximate unlearning approaches that efficiently minimize data influence. By comprehensively reviewing solutions, we identify and discuss their strengths and limitations. Furthermore, we propose future directions to advance machine unlearning and establish it as an essential capability for trustworthy and adaptive machine learning models. This paper provides researchers with a roadmap of open problems, encouraging impactful contributions to address real-world needs for selective data removal.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"8 3","pages":"2150-2168"},"PeriodicalIF":5.3,"publicationDate":"2024-04-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141096368","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-04-02DOI: 10.1109/TETCI.2024.3380442
Sucheng Ren;Nanxuan Zhao;Qiang Wen;Guoqiang Han;Shengfeng He
The fully convolutional network (FCN) has dominated salient object detection for a long period. However, the locality of CNN requires the model deep enough to have a global receptive field and such a deep model always leads to the loss of local details. In this paper, we introduce a new attention-based encoder, vision transformer, into salient object detection to ensure the globalization of the representations from shallow to deep layers. With the global view in very shallow layers, the transformer encoder preserves more local representations to recover the spatial details in final saliency maps. Besides, as each layer can capture a global view of its previous layer, adjacent layers can implicitly maximize the representation differences and minimize the redundant features, making every output feature of transformer layers contribute uniquely to the final prediction. To decode features from the transformer, we propose a simple yet effective deeply-transformed decoder. The decoder densely decodes and upsamples the transformer features, generating the final saliency map with less noise injection. Experimental results demonstrate that our method significantly outperforms other FCN-based and transformer-based methods in five benchmarks by a large margin, with an average of 12.17% improvement in terms of Mean Absolute Error (MAE).
{"title":"Unifying Global-Local Representations in Salient Object Detection With Transformers","authors":"Sucheng Ren;Nanxuan Zhao;Qiang Wen;Guoqiang Han;Shengfeng He","doi":"10.1109/TETCI.2024.3380442","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3380442","url":null,"abstract":"The fully convolutional network (FCN) has dominated salient object detection for a long period. However, the locality of CNN requires the model deep enough to have a global receptive field and such a deep model always leads to the loss of local details. In this paper, we introduce a new attention-based encoder, vision transformer, into salient object detection to ensure the globalization of the representations from shallow to deep layers. With the global view in very shallow layers, the transformer encoder preserves more local representations to recover the spatial details in final saliency maps. Besides, as each layer can capture a global view of its previous layer, adjacent layers can implicitly maximize the representation differences and minimize the redundant features, making every output feature of transformer layers contribute uniquely to the final prediction. To decode features from the transformer, we propose a simple yet effective deeply-transformed decoder. The decoder densely decodes and upsamples the transformer features, generating the final saliency map with less noise injection. Experimental results demonstrate that our method significantly outperforms other FCN-based and transformer-based methods in five benchmarks by a large margin, with an average of 12.17% improvement in terms of Mean Absolute Error (MAE).","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"8 4","pages":"2870-2879"},"PeriodicalIF":5.3,"publicationDate":"2024-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141965864","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-04-02DOI: 10.1109/TETCI.2024.3377267
Xuejiao Li;Jun Chen;Heye Zhang;Yongwon Cho;Sung Ho Hwang;Zhifan Gao;Guang Yang
Three-dimensional left atrial (LA) segmentation from late gadolinium-enhanced cardiac magnetic resonance (LGE CMR) images is of great significance in the prevention and treatment of atrial fibrillation. Despite deep learning-based approaches have made significant progress in 3D LA segmentation, they usually require a large number of labeled images for training. Few-shot learning can quickly adapt to novel tasks with only a few data samples. However, the resolution discrepancy of LGE CMR images presents challenges for few-shot learning in 3D LA segmentation. To address this issue, we propose the Hierarchical Relational Inference Network (HRIN), which extracts the interactive features of support and query volumes through a bidirectional hierarchical relationship learning module. HRIN learns the commonality and discrepancy between support and query volumes by modeling the higher-order relations. Notably, we embed the bidirectional interaction information between support and query volumes into the prototypes to adaptively predict the query. Additionally, we leverage prior knowledge of foreground and background information in the support volume to model queries. We validated the performance of our method on a total of 369 scans from two centers. Our proposed HRIN achieves higher segmentation performance compared to other state-of-the-art segmentation methods. With only 5% data samples, the average Dice Similarity Coefficient of the two centers respectively reaches 0.8454 and 0.8110. Compared with other methods under the same conditions, the highest values only reach 0.7012 and 0.6898. Our approach improves the adaptability and generalization of few-shot segmentation from LGE CMR images, enabling precise evaluation of LA remodeling.
从晚期钆增强心脏磁共振（LGE CMR）图像中进行三维左心房（LA）分割对预防和治疗心房颤动具有重要意义。尽管基于深度学习的方法在三维 LA 分割方面取得了重大进展，但它们通常需要大量标记图像进行训练。少量学习只需少量数据样本就能快速适应新任务。然而，LGE CMR 图像的分辨率差异给三维 LA 分割中的少量学习带来了挑战。为解决这一问题，我们提出了层次关系推理网络（HRIN），通过双向层次关系学习模块提取支持量和查询量的交互特征。HRIN 通过对高阶关系建模来学习支持量和查询量之间的共性和差异。值得注意的是，我们将支持量和查询量之间的双向交互信息嵌入到原型中，以便自适应地预测查询。此外，我们还利用支持卷中的前景和背景信息的先验知识对查询进行建模。我们在两个中心的总共 369 次扫描中验证了我们方法的性能。与其他最先进的分割方法相比，我们提出的 HRIN 实现了更高的分割性能。在只有 5%数据样本的情况下，两个中心的平均骰子相似系数分别达到了 0.8454 和 0.8110。与相同条件下的其他方法相比，最高值仅为 0.7012 和 0.6898。我们的方法提高了从 LGE CMR 图像中进行少次分割的适应性和通用性，从而能够精确评估 LA 重塑情况。
{"title":"Hierarchical Relational Inference for Few-Shot Learning in 3D Left Atrial Segmentation","authors":"Xuejiao Li;Jun Chen;Heye Zhang;Yongwon Cho;Sung Ho Hwang;Zhifan Gao;Guang Yang","doi":"10.1109/TETCI.2024.3377267","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3377267","url":null,"abstract":"Three-dimensional left atrial (LA) segmentation from late gadolinium-enhanced cardiac magnetic resonance (LGE CMR) images is of great significance in the prevention and treatment of atrial fibrillation. Despite deep learning-based approaches have made significant progress in 3D LA segmentation, they usually require a large number of labeled images for training. Few-shot learning can quickly adapt to novel tasks with only a few data samples. However, the resolution discrepancy of LGE CMR images presents challenges for few-shot learning in 3D LA segmentation. To address this issue, we propose the Hierarchical Relational Inference Network (HRIN), which extracts the interactive features of support and query volumes through a bidirectional hierarchical relationship learning module. HRIN learns the commonality and discrepancy between support and query volumes by modeling the higher-order relations. Notably, we embed the bidirectional interaction information between support and query volumes into the prototypes to adaptively predict the query. Additionally, we leverage prior knowledge of foreground and background information in the support volume to model queries. We validated the performance of our method on a total of 369 scans from two centers. Our proposed HRIN achieves higher segmentation performance compared to other state-of-the-art segmentation methods. With only 5% data samples, the average Dice Similarity Coefficient of the two centers respectively reaches 0.8454 and 0.8110. Compared with other methods under the same conditions, the highest values only reach 0.7012 and 0.6898. Our approach improves the adaptability and generalization of few-shot segmentation from LGE CMR images, enabling precise evaluation of LA remodeling.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"8 5","pages":"3352-3367"},"PeriodicalIF":5.3,"publicationDate":"2024-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142368450","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-04-02DOI: 10.1109/TETCI.2024.3377676
Mohammed Yusuf Ansari;Iffa Afsa Changaai Mangalote;Pramod Kumar Meher;Omar Aboumarzouk;Abdulla Al-Ansari;Osama Halabi;Sarada Prasad Dakua
Ultrasound (US) is generally preferred because it is of low-cost, safe, and non-invasive. US image segmentation is crucial in image analysis. Recently, deep learning-based methods are increasingly being used to segment US images. This survey systematically summarizes and highlights crucial aspects of the deep learning techniques developed in the last five years for US segmentation of various body regions. We investigate and analyze the most popular loss functions and metrics for training and evaluating the neural network for US segmentation. Furthermore, we study the patterns in neural network architectures proposed for the segmentation of various regions of interest. We present neural network modules and priors that address the anatomical challenges associated with different body organs in US images. We have found that variants of U-Net that have dedicated modules to overcome the low-contrast and blurry nature of images are suitable for US image segmentation. Finally, we also discuss the advantages and challenges associated with deep learning methods in the context of US image segmentation.
超声波（US）因其成本低、安全、无创伤而受到普遍青睐。US 图像分割在图像分析中至关重要。最近，基于深度学习的方法越来越多地被用于 US 图像分割。本调查系统地总结并强调了过去五年中开发的深度学习技术的关键方面，这些技术用于对不同身体区域的 US 图像进行分割。我们研究并分析了最流行的损失函数和指标，用于训练和评估用于 US 分割的神经网络。此外，我们还研究了为分割各种感兴趣区域而提出的神经网络架构的模式。我们提出了神经网络模块和先验，以应对 US 图像中与不同人体器官相关的解剖学挑战。我们发现，具有专用模块以克服图像低对比度和模糊特性的 U-Net 变体适用于 US 图像分割。最后，我们还讨论了深度学习方法在 US 图像分割方面的优势和挑战。
{"title":"Advancements in Deep Learning for B-Mode Ultrasound Segmentation: A Comprehensive Review","authors":"Mohammed Yusuf Ansari;Iffa Afsa Changaai Mangalote;Pramod Kumar Meher;Omar Aboumarzouk;Abdulla Al-Ansari;Osama Halabi;Sarada Prasad Dakua","doi":"10.1109/TETCI.2024.3377676","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3377676","url":null,"abstract":"Ultrasound (US) is generally preferred because it is of low-cost, safe, and non-invasive. US image segmentation is crucial in image analysis. Recently, deep learning-based methods are increasingly being used to segment US images. This survey systematically summarizes and highlights crucial aspects of the deep learning techniques developed in the last five years for US segmentation of various body regions. We investigate and analyze the most popular loss functions and metrics for training and evaluating the neural network for US segmentation. Furthermore, we study the patterns in neural network architectures proposed for the segmentation of various regions of interest. We present neural network modules and priors that address the anatomical challenges associated with different body organs in US images. We have found that variants of U-Net that have dedicated modules to overcome the low-contrast and blurry nature of images are suitable for US image segmentation. Finally, we also discuss the advantages and challenges associated with deep learning methods in the context of US image segmentation.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"8 3","pages":"2126-2149"},"PeriodicalIF":5.3,"publicationDate":"2024-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141095529","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The functional connections in the human brain offer many opportunities to explore changing dynamic patterns of the brain under different circumstances. Different factors such as age, mental activity, and health status may affect functional connectivity, connected regions, and the robustness of connections in the brain. In this study, we evaluate the functional connectivity of the whole brain changing with age from a complex network perspective during different processes in healthy adults. We conducted a functional Magnetic Resonance Imaging (fMRI) study that includes both resting and cognitive states with elderly and young participants (n = 38). To analyze the functional connectivity structure in view of graph theory, we used the minimum dominating sets (MDS) and then minimum hitting sets (MHS) of the connectivity networks. Based on our analysis, age, and mental activity show a significant effect on the hitting sets and dominating sets of the brain regions. The results also indicate that the working mechanism of the brain changes from local to diffused under the circumstances of a particular computational load with age. In this manner, the proposed method can be used as a complementary method for clinical procedures to evaluate and measure the effect of aging on the human brain.
{"title":"The Impact of Mental Activities and Age on Brain Network: An Analysis From Complex Network Perspective","authors":"Cemre Candemir;Vahid Khalilpour Akram;Ali Saffet Gonul","doi":"10.1109/TETCI.2024.3374957","DOIUrl":"https://doi.org/10.1109/TETCI.2024.3374957","url":null,"abstract":"The functional connections in the human brain offer many opportunities to explore changing dynamic patterns of the brain under different circumstances. Different factors such as age, mental activity, and health status may affect functional connectivity, connected regions, and the robustness of connections in the brain. In this study, we evaluate the functional connectivity of the whole brain changing with age from a complex network perspective during different processes in healthy adults. We conducted a functional Magnetic Resonance Imaging (fMRI) study that includes both resting and cognitive states with elderly and young participants (n = 38). To analyze the functional connectivity structure in view of graph theory, we used the minimum dominating sets (MDS) and then minimum hitting sets (MHS) of the connectivity networks. Based on our analysis, age, and mental activity show a significant effect on the hitting sets and dominating sets of the brain regions. The results also indicate that the working mechanism of the brain changes from local to diffused under the circumstances of a particular computational load with age. In this manner, the proposed method can be used as a complementary method for clinical procedures to evaluate and measure the effect of aging on the human brain.","PeriodicalId":13135,"journal":{"name":"IEEE Transactions on Emerging Topics in Computational Intelligence","volume":"8 4","pages":"2791-2803"},"PeriodicalIF":5.3,"publicationDate":"2024-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141965841","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：