检索规则说明:AND代表“并且”;OR代表“或者”;NOT代表“不包含”;(注意必须大写,运算符两边需空一格)
检 索 范 例 :范例一: (K=图书馆学 OR K=情报学) AND A=范并思 范例二:J=计算机应用与软件 AND (U=C++ OR U=Basic) NOT M=Visual
作 者:杨文琴[1] YANG Wenqin(The Academy of Digital China(Fujian),Fuzhou University,Fuzhou 350108,China)
机构地区:[1]福州大学数字中国研究院(福建),福建福州350108
出 处:《长江信息通信》2023年第12期1-5,9,共6页Changjiang Information & Communications
基 金:福建省科技计划:中央引导地方科技发展专项项目(No.2022L3003)。
摘 要:为保护多方医院数据隐私以及数据不平衡等问题,以肾脏肿瘤CT影像为研究对象,提出一种基于个性化联邦学习的肾脏肿瘤分割方法。首先,以经典的深度残差网络SEResNet为基础,结合深监督模块来提高目标分割的精度;其次,在联邦平均算法框架下,引入基础层与个性化层联合学习策略,克服多方数据不平衡带来模型精度下降问题。经KiTS21挑战赛数据集实验验证,文章方法获得的肾脏+肿瘤+囊肿、肿瘤+囊肿和肿瘤分割Dice分别为93.61%、67.18%、61.01%,与集中式学习相比,仅分别低0.56%、7.18%和9.34%,表明本文方法以可接受的精度损失换取了数据的隐私安全。In order to protect data privacy and data imbalance in multiple hospitals,this paper proposes a personalized federated learning-based method for renal tumor segmentation based on CT images.Firstly,based on the classical deep residual network SEResNet,the method incorporates deep supervision to improve the segmentation accuracy.Then,the federated averaging algorithm framework is adopted,and a joint learning method with both basic and personalized layer is designed to overcome the problem of model accuracy degradation caused by multi-party data imbalance.Experiment in the KiTS21 Challenge dataset,the proposed method achieves Dice scores of 93.61%,67.18%,61.01% for kidney+tumor+cyst,tumor+cyst,and tumor segmentation,respectively.Compared with the results of centralized learning,the results are only 0.56%,7.18%,and 9.34%lower.The results indicate that the method proposed can approach the results of centralized learning with an acceptable loss of accuracy,while ensuring data security.
分 类 号:TP391[自动化与计算机技术—计算机应用技术]
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在载入数据...
正在链接到云南高校图书馆文献保障联盟下载...
云南高校图书馆联盟文献共享服务平台 版权所有©
您的IP:216.73.216.124