Efficient Simultaneous Reconstruction of Time-Varying Images and Electrode Contact Impedances in Electrical Impedance Tomography

• Published in: IEEE transactions on biomedical engineering 2017-04, Vol.64 (4), p.795-806
• Main Authors: Boverman, Gregory, Isaacson, David, Newell, Jonathan C., Saulnier, Gary J., Kao, Tzu-Jen, Amm, Bruce C., Wang, Xin, Davenport, David M., Chong, David H., Sahni, Rakesh, Ashe, Jeffrey M.
• Format: Article
• Language: English
• Subjects: Algorithms; Bioimpedance; biomedical imaging; Conductivity; Current measurement; Electric contacts; Electric Impedance; Electrical impedance; Electrical resistivity; Electrodes; Equipment Design; Equipment Failure Analysis; Humans; Image Interpretation, Computer-Assisted - methods; Image reconstruction; Impedance; Linear algebra; Linearization; Mathematical analysis; Mathematical models; Parameters; Permittivity; Plethysmography, Impedance - instrumentation; Plethysmography, Impedance - methods; reconstruction algorithms; Regularization; Reproducibility of Results; Sensitivity; Sensitivity and Specificity; Singular value decomposition; Tomography; Tomography - instrumentation; Tomography - methods
• ISSN: 0018-9294; 1558-2531
• DOI: 10.1109/TBME.2016.2578646

Objective: In electrical impedance tomography (EIT), we apply patterns of currents on a set of electrodes at the external boundary of an object, measure the resulting potentials at the electrodes, and, given the aggregate dataset, reconstruct the complex conductivity and permittivity within the object. It is possible to maximize sensitivity to internal conductivity changes by simultaneously applying currents and measuring potentials on all electrodes but this approach also maximizes sensitivity to changes in impedance at the interface. Methods: We have, therefore, developed algorithms to assess contact impedance changes at the interface as well as to efficiently and simultaneously reconstruct internal conductivity/permittivity changes within the body. We use simple linear algebraic manipulations, the generalized singular value decomposition, and a dual-mesh finite-element-based framework to reconstruct images in real time. We are also able to efficiently compute the linearized reconstruction for a wide range of regularization parameters and to compute both the generalized cross-validation parameter as well as the L-curve, objective approaches to determining the optimal regularization parameter, in a similarly efficient manner. Results: Results are shown using data from a normal subject and from a clinical intensive care unit patient, both acquired with the GE GENESIS prototype EIT system, demonstrating significantly reduced boundary artifacts due to electrode drift and motion artifact.