Electrical Impedance Tomography (EIT) is an imaging modality in which electrical conductivity within the object is estimated from surface voltage measurements. Conductivity images in EIT can provide clinically significant information, since conductivity changes are closely related with physiological changes inside the body. Clinical applications of EIT are non-invasive, cost-effective and simple to apply to patients. Despite the above advantages, performance of EIT systems is affected by the uncertainties pertaining to patient’s body shape change and spatial variability of the image reconstruction problem. In this research, we identify these uncertainties to be the major sources of reconstruction errors.

In EIT image reconstruction, the surface shape of an image object is often assumed to be known. In clinical environments, shape information is not always available. Discrepancies between the assumed and actual shapes can result in errors that may have clinical significance. We suggest an algorithm that estimates domain shapes for the use in 2D EIT. We investigated elliptical boundary distortions of a unit disk object as changes from circular to elliptical geometries, defined using the Joukowski transformation. Boundary shapes of a real domain were then estimated as ellipses after investigating the spatial characteristics of image artifacts caused by shape changes. Our method was tested with boundary voltage measurements obtained using a full array electrode layout from elliptical simulation and phantom models containing a small disk anomaly at various positions. We found that the proposed method could estimate elliptical shape changes with relatively small error.

The EIT image reconstruction problem is spatially-variant, meaning that the same anomaly placed at different locations within an image plane may produce different reconstruction signatures. Correcting errors due to this spatial variability should improve reconstruction accuracy. We present methods to normalize the spatially-variant image reconstruction problem by equalizing the system Point Spread Function (PSF). In order to equalize PSF, we used blurring properties of the system derived from the sensitivity matrix. We compared three mathematical schemes: Pixel-Wise Scaling (PWS), Weighted Pseduo-Inversion (WPI) and Weighted Minimum Norm Method (WMNM) to normalize reconstructions. The Quantity Index (QI), defined as the integral of pixel values of an EIT conductivity image, was considered in investigating spatial variability. The QI values along with reconstructed images are presented for cases of 2D full array and hemiarray electrode topologies. We found that a less spatially-variant QI could be obtained by applying normalization methods to conventional regularized reconstruction methods such as Truncated Singular Value Decomposition (TSVD) and WMNM. The normalization methods were tested with boundary voltage measurements obtained from simulation disk models containing a smaller disk anomaly, and cylindrical phantom models with anomalies of various volumes placed at various locations within the electrode plane. For anomalies of the same volume, QI error caused by spatial variability was reduced the most among the tested methods when WMNM normalization was applied to WMNM regularized reconstructions for both hemiarray and full array cases.

The use of the blurring properties was further investigated in hemiarray EIT, where the electrodes cover only one half of the object boundary. Boundary measurements are relatively not sensitive to the conductivity anomaly that lies far away from electrodes, and the anomaly may be invisible or undetected in the images reconstructed using conventional methods. We propose a WPI method to enhance sensitivity in the region distant from the electrodes. The method was tested with data obtained from a 2D circular object. A smaller disk anomaly was varied in location within the object. The WPI method detected anomalies with relatively small errors for the hemiarray case.