Linearly constrained minimum variance spatial filtering for localization of conductivity changes in electrical impedance tomography

• Published in: International journal for numerical methods in biomedical engineering 2015-02, Vol.31 (2), p.np-n/a
• Main Authors: Fernández-Corazza, M., von Ellenrieder, N., Muravchik, C. H.
• Format: Article
• Language: English
• Subjects: Algorithms; Computer Simulation; Constraints; Electric Impedance; Electrical impedance; electrical impedance tomography; Humans; linearly constrained minimum variance spatial filter; Localization; localization of conductivity changes; Phantoms, Imaging; Position (location); Reconstruction; Resistivity; Spatial filtering; Tomography; Tomography - methods
• ISSN: 2040-7939; 2040-7947
• DOI: 10.1002/cnm.2703

SummaryWe localize dynamic electrical conductivity changes and reconstruct their time evolution introducing the spatial filtering technique to electrical impedance tomography (EIT). More precisely, we use the unit‐noise‐gain constrained variation of the distortionless‐response linearly constrained minimum variance spatial filter. We address the effects of interference and the use of zero gain constraints. The approach is successfully tested in simulated and real tank phantoms. We compute the position error and resolution to compare the localization performance of the proposed method with the one‐step Gauss–Newton reconstruction with Laplacian prior. We also study the effects of sensor position errors. Our results show that EIT spatial filtering is useful for localizing conductivity changes of relatively small size and for estimating their time‐courses. Some potential dynamic EIT applications such as acute ischemic stroke detection and neuronal activity localization may benefit from the higher resolution of spatial filters as compared to conventional tomographic reconstruction algorithms. Copyright © 2015 John Wiley & Sons, Ltd. We introduce the use of the linearly constrained minimum variance spatial filter to localize conductivity changes and to estimate its time courses from simulated and phantom‐generated electrical impedance tomography measurements. Results show that it performs better in terms of bias and resolution than a standard electrical impedance tomography reconstruction algorithm. The technique is promising in biomedical applications such as ischemic stroke detection or neuronal activity localization.