期刊文献+

从U-Net到Transformer:深度模型在医学图像分割中的应用综述

From U-Net to Transformer:application review of deep models in medical image segmentation
下载PDF
导出
摘要 精准分割医学图像中的病灶对医生探寻病因和制定诊疗方案起关键作用,计算机视觉技术的发展促使深度学习在医学图像分割领域衍生出多种模型架构。U-Net架构以其巧妙的跳跃连接、易于优化的模块设计成为这一领域的基准模型。然而,U-Net以卷积神经网络(CNN)为主干,在长期建模依赖关系方面只擅长获取局部特征,基于CNN的各项方法在执行分割任务中缺乏对图像长期相关性的解释,无法提取全局特征。为帮助本领域学者了解U-Net的发展历程及研究现状,以问题为导向对2016-2023年U-Net改进工作进行综述。首先,从改进结构位置的角度对U-Net及其各项改进模型进行叙述,探讨各工作的研究目的和创新设计及不足之处;其次,对Transformer与U-Net的结合方式进行分析,从中获取改进工作的研究动向;最后,在Synapse和ACDC数据集上进行对比实验,通过实验分析和可视化结果表明,Transformer方法在分割精度方面有显著优势,特别是混合网络子块的结合方式,在确保模型性能的同时兼顾效率,证明了该类工作有着广阔的发展前景和研究价值。 The accurate segmentation of lesions in medical images plays a key role in the physician’s search for the cause of disease and the formulation of treatment plans.The development of computer vision technology has led to the derivation of various model architectures for deep learning in medical image segmentation.The U-Net(U-shaped Network)architecture has become a benchmark model in this field for its clever skip connection and easy-to-optimize module design.However,U-Net network,with Convolutional Neural Network(CNN)as its backbone,is only good at acquiring local features in terms of long-term modeling dependencies,and various CNN-based methods lack the interpretation of long-term image correlations in performing segmentation tasks to extract global features.In order to help scholars in this field to understand the development history and research status of U-Net,a problem-oriented overview of the U-Net improvement work between 2016 and 2023 was presented.Firstly,the U-Net and its various improved models were narrated from the perspective of improving the structural position,and the research purpose,innovative design and shortcomings of each work were discussed.Secondly,the combination method of Transformer and U-Net was analyzed to obtain research trends for improvement work.Finally,comparative experiments were conducted on Synapse and ACDC datasets,and the experimental analysis and visualization results show that the Transformer-based method has a significant advantage in segmentation accuracy,especially the way of combining hybrid network sub-blocks,which ensures the model performance while taking into account the efficiency,and proves that this class of work has broad development prospect and research value.
作者 张玮智 于谦 苏金善 乎西旦·居马洪 林玲 ZHANG Weizhi;YU Qian;SU Jinshan;HUXIDAN Jumahong;LIN Ling(School of Network Security and Information Technology,Yili Normal University,Yining Xinjiang 835000,China;Key Laboratory of Intelligent Computing Research and Application,Yili Normal University,Yining Xinjiang 835000,China;School of Data and Computer Science,Shandong Women’s University,Jinan Shandong 255300,China;School of Electronic and Engineering,Yili Normal University,Yining Xinjiang 835000,China;Key Laboratory of Vibration Signal Capture and Intelligent Processing,Yili Normal University,Yining Xinjiang 835000,China)
出处 《计算机应用》 CSCD 北大核心 2024年第S01期204-222,共19页 journal of Computer Applications
基金 伊犁师范大学校级重点项目(2023YSZD006) 国家自然科学基金资助项目(62266046) 山东省自然科学基金资助项目(ZR2023MF037) 新疆维吾尔自治区研究生创新项目(XJ2023G258)。
关键词 医学图像分割 U-Net 结构改进 TRANSFORMER 深度神经网络 medical image segmentation U-Net structural improvement Transformer deep neural network
  • 相关文献

参考文献2

二级参考文献25

共引文献33

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部