Dynamic electrical impedance tomography imaging algorithm based on complementary information fusion network
10.19745/j.1003-8868.2025098
- VernacularTitle:基于互补信息融合网络的动态电阻抗断层成像算法研究
- Author:
Xin-yi WANG
1
;
Tao ZHANG
;
Xiang TIAN
;
Ning YANG
;
Jun-jie DU
;
Xue-chao LIU
;
Feng FU
;
Xue-tao SHI
;
Can-hua XU
Author Information
1. 空军军医大学军事生物医学工程学系,陕西省生物电磁检测与智能感知重点实验室,西安 710032
- Publication Type:Journal Article
- Keywords:
electrical impedance tomography imaging;
complementary information fusion network;
deep learning;
image reconstruction
- From:
Chinese Medical Equipment Journal
2025;46(6):1-6
- CountryChina
- Language:Chinese
-
Abstract:
Objective To propose a dynamic electrical impedance tomography imaging algorithm based on complementary information fusion network(CIFN)to enhance image quality of dynamic electrical impedance imaging.Methods There were three modules for initialization,multi-frame complementary information extraction and information fusion involved in the CIFN.Firstly,multi-frame dynamic conductivity distribution images were obtained by the initialization module;secondly,spatial complementary information was extracted from the images by using the multi-frame complementary information extraction module;finally,the fusion of lesion target distribution information and target re-reconstruction were realized by the information fusion module to aquire high-quality EIT images.With a 16-electrode multilayer cranial simulation model,the CIFN-based imaging method was compared with Tikhonov regularization algorithm,spectral constraint algorithm and U-Net algorithm in terms of imaging results of types of lesions to verify its performance.Results Compared with the Tikhonov regularization algorithm,spectral constraint algorithm and U-Net algorithm,the proposed CIFN-based algorithm exhibited the lowest mean absolute error(MAE)and the highest structural similarity(SSIM)when used to image different lesion targets,which accurately reconstructed the distribution of lesion targets and gained high imaging stability under common noise levels.Conclusion The proposed CIFN-based imaging algorithm obtains high imaging quality on a cranial simulation model and reconstruction results close to the real model distribution,which provides algorithmic support for subsequent clinical studies on electrical impedance imaging.[Chinese Medical Equipment Journal,2025,46(6):1-6]
