An Improved Conjugate Gradient Image Reconstruction Algorithm for Electromagnetic Tomography

• Published in: Sensing and imaging Vol. 23; no. 1
• Main Authors: Liu, Xianglong, Wang, Ying
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.12.2022; Springer Nature B.V
• Subjects: Algorithms; Biomedical materials; Conjugate gradient method; Correlation coefficients; Electrical Engineering; Engineering; Ill posed problems; Image enhancement; Image quality; Image reconstruction; Imaging; Inverse problems; Iterative algorithms; Iterative methods; Microwaves; Original Paper; Radiology; Regularization; Regularization methods; RF and Optical Engineering; Spatial resolution; Tomography; Improved conjugate gradient; Image reconstruction; Electromagnetic tomography; Inverse problem
• ISSN: 1557-2064; 1557-2072
• DOI: 10.1007/s11220-021-00374-y

Electromagnetic tomography (EMT) is an emerging imaging modality capable of visualizing the distribution of electrically conductive or magnetically permeable materials within the vessels and pipelines. Image reconstruction is the crucial step of EMT inverse problem, which is one of the main challenges in the promotion and application of EMT to industrial and biomedical fields. EMT inverse problem is ill-posed and ill-conditioned, which is the key to the reconstructed images quality. Tikhonov regularization is the widely used regularization method to solve the ill-posed problem. The revised Tikhonov regularization is obtained by improving the Tikhonov regularization when observation noise is considered. This paper presents an improved conjugate gradient algorithm based on the revised Tikhonov regularization to reduce the ill-posed nature of EMT inverse problem and enhance the spatial resolution of the reconstructed images. Numerical simulations results confirm that the quality of the reconstructed images obtained by the proposed algorithm is improved and also better than the other conventional algorithms including linear back projection, Tikhonov regularization algorithm, Landweber iterative algorithm, in terms of the typical patterns. Besides, correlation coefficients of the proposed ICGRT algorithm are 0.5961, 0.5989 and 0.5231 for three experimental typical patterns. The proposed ICGRT algorithm has highest correlation coefficients and reconstructs the best images.