Assessing the effect of error in modelling on accuracy of EIT image reconstruction

EIT is a simple and robust imaging technique that is vastly applied for both medical and industrial process imaging. One of the biggest influencing factors in deciding the accuracy of the reconstructed image lies in the model. As EIT is an underdetermined, nonlinear system, if the number of elements in a discretized model that needs to be solved becomes too large, and the number of available measurements are limited, this will result in inaccuracy in the reconstructed images. This tends to become an issue when the model is a complex one, which tends to be the case in medical imaging, and a large number of elements are required to sufficiently represent the details in the model. One way of reducing the complication in computation is simplifying the model through stripping segments that do not contribute directly to the region of interest, or simply by not including the details in the model. This, however, would result in error in the reconstructed image. This paper presents an illustration of the importance of accuracy in modelling, and the effect of discrepancies in model on the overall interpretation of results. A comparative study for reconstruction obtained using different number of electrodes is used as a platform for this illustration. For the first study, results are obtained using a single model, but configured differently. For the second study, the results are obtained using models set-up separately, generated based on the basic requirement of the electrodes set-up. The results of both studies are used to illustrate the importance of accuracy in modelling and its effect on the reconstructed images. Results indicate that when the level of discretization of the models is significant (approximately above 40%) the results become noticeably different.