Performance evaluation of image reconstruction algorithms in an electromagnetic tomographic system

The aim of electromagnetic tomography (EMT) or magnetic induction tomography is to estimate the electrical conductivity σ and the magnetic permeability μ distributions inside a given region of interest (ROI), by measuring the induced voltage at a number of detection coils placed around it. The induced voltages are nonlinearly related to the materials´ electrical and magnetic properties since the magnetic flux paths are distorted by the unknown distributions. Therefore, EMT (such as EIT) is a nonlinear inverse problem that requires far more complex and computer intensive algorithms for image reconstruction than, for instance, X-ray transmission tomography which is linear. However, if the distribution of materials inside the ROI consists of a small group of homogeneous objects which are separated by distances at least as large as their size, object interaction is not very strong and the nonlinear effects are minimised. For these cases, very fast algorithms which produce qualitative images can be used. Here, a performance evaluation of 3 of those algorithms is carried out. The presented results are based on real data which was collected on a EMT system developed at INESC (Aveiro, 1996). It is concluded that, if object interaction is not very strong, the inverse problem in electromagnetic tomography can be dealt with in a linear way (ART), or approximated to a linear problem (1 iteration Newton-Raphson). Qualitative images can then be generated in a very fast and efficient way which is very important for some industrial applications where real time imaging is required