摘要
电阻抗层析成像技术(EIT)是一种非侵入式且无辐射的成像方法,备受医学领域关注。由于EIT问题的病态,不适定性及非线性特性,导致其重建图像分辨率低,伪影严重。为了提高EIT成像的空间分辨率,本文提出一种堆栈式自编码器(EIT-SAE)成像方法,采用监督式学习,构建肺部EIT图像数据库,经过训练学习,使自编码器具有良好的非线性表征能力,实现从边界电压测量值到场域电导率分布的映射,与正则化和共轭梯度算法相比,本文提出的自编码器重建方法具有良好的重建方法能够大幅减少伪影,成像质量得到显著提升。
Electrical impedance tomography(EIT) is a non-invasive and non-radiation imaging method that attracts the attention in the medical field. However, due to the ill-posed and nonlinear characterizations of electrical impedance imaging problem, the reconstructed image has a low spatial resolution with artifacts. Aiming to improve the imaging spatial quality, this paper proposes a new imaging method on stacked autoencoder(SAE).It is studied with supervisory, and we also make a special lung EIT database. After training and test, the network has strong capability for nonlinear mapping from the boundary voltages to the conductivities distribution. Compared with regularization and conjugate gradient algorithms, the proposed method has good effect on EIT image with less artifacts and the quality improved well.
作者
付荣
张新宇
王子辰
王迪
陈晓艳
张淼
FU Rong;ZHANG Xinyu;WANG Zichen;WANG Di;CHEN Xiaoyan;ZHANG Miao(College of Electronic Ilnformation and Automation,Tianjin University of cience&Technology,Tianjin,300202)
出处
《生命科学仪器》
2021年第1期93-98,共6页
Life Science Instruments
基金
国家自然科学基金资助(61903724,41704131)
天津市自然科学基金资助(18YFZCGX00360)。
