一种基于深度学习的2.5D心脏分割算法  

2.5D cardiac segmentation algorithm based on deep learning

作　　者：徐佳陈 龙翔 Xu Jiachen;Long Xiang(School of Computer and Information,Nantong Institute of Technology,Nantong 226001,China)

机构地区：[1]南通理工学院计算机与信息工程学院,江苏南通226001

出　　处：《无线互联科技》2024年第10期25-27,60,共4页Wireless Internet Technology

基　　金：2023年度南通理工学院科研项目;项目名称:基于k-匿名的轨迹隐私保护方法研究;项目编号:2022XK(Z)27。

摘　　要：心脏动态核磁共振图像的自动分割算法可以有效减少心血管科医生压力,其分割结果是评估心脏功能的金标准。文章提出了一种有效分割心脏动态核磁共振图像的自动分割算法。该算法采用U形神经网络结果,根据心脏动态核磁共振图像,网络结构分为目标图像下采样层、相邻时间片下采样层、上采样层和连接桥。与其他分割算法相比,文章提出的算法能够更好地分割心脏图像,分割效果更为精准。The automatic segmentation algorithm of cardiac dynamic MRI images can effectively reduce the pressure on cardiovascular doctors,and its segmentation results are the gold standard for evaluating cardiac function.This article proposes an automatic segmentation algorithm that effectively segments cardiac dynamic MRI images.According to the cardiac dynamic MRI image,the U-shaped neural network results are used by the proposed algorithm.The network structure is divided into a target image in the downsampling layer,an adjacent time slice in the downsampling layer,an upsampling layer and a connecting bridge.Compared with other segmentation algorithms,the proposed algorithm in this article can better segment cardiac images and the segmentation effect is more accurate.

关 键 词：图像分割 心脏动态核磁共振图像 Unet 医学影像 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]