摘要
电阻抗成像图像重建是一个高度非线性和不适定性问题,为解决传统图像重建方法丢失信息、无法实现高精度和实时成像的问题,文章提出了一种基于LeNet-5卷积神经网络结构的电阻抗图像重建方法。该方法利用MATLAB以及COMSOL联合仿真建立规则的圆形模型以及符合人体几何特异性的肺部模型,生成具有不同成像特征的样本,这些样本之后被划分为训练集、验证集以及测试集,经过LeNet-5网络学习边界测量电压值和电导率之间的非线性关系从而实现图像重建。将文中网络获取的重建结果与其他机器学习方法(BP神经网络、RBF神经网络)的结果进行比较,验证基于LeNet-5的网络应用于图像重建的有效性。
The image reconstruction of electrical impedance tomography is a highly nonlinear and ill-posed problem,and the traditional image reconstruction method loses important information,which means it cannot achieve high accuracy and real-time imaging.In this paper,an electrical impedance tomography image reconstruction method based on convolutional ncural network structure LeNet-5 is proposed.MATLAB and COMSOL were used to obtain samples with dif-ferent representative imaging features,and these samples were divided into training set,valida-tion set and test set.The LeNet-5 network is used to learn the nonlinear relationship between the boundary measurement voltage and conductivity to reconstruct images.The reconstruction re-sults obtained by the network in this paper were compared with the results obtained by other machine learning methods(backpropagation and radial basis function neural networks)to verify the effectiveness of the network based on LeNet-5.
作者
李少聪
张清河
郑国亮
LI Shaocong;ZHANG Qinghe;ZHENG Guoliang(Hubei Key Laboratory of Intelligent Vision Based Monitoring for Hydroelectric Engineering,China Three Gorges University,Yichang 443002,China;School of computer and infomation,China Three Gorges University,Yichang Hubei 443002,China)
出处
《长江信息通信》
2024年第5期78-82,共5页
Changjiang Information & Communications
关键词
电阻抗层析成像
卷积神经网络
深度学习
图像重建
clectrical impedance tomography
convolutional neural network
deep learning
im-age rcconstruction
