Variations in in vivo electrical impedance tomography images due to inaccuracy in boundary representation

A comparison of in vivo image results is performed for five image-reconstruction programs, featuring an increase in accuracy of boundary modelling from a simple 2-D disk to a true boundary shape with each current drive field individually calculated. Variations are found both in the positions of imaged features and their appearance, but reasonable consistency in reconstructed impedance changes is obtained for both phantom and in vivo data. In terms of quantitative measurements, the programs based on the simpler boundary assumptions generally perform more reliably than the more complex versions. It is concluded that the quantitative use of EIT with simple boundary assumptions is not compromised by body contour variations between patients, provided that the appropriate regions of interest can be correctly identified.