Molinari, M, Cox, SJ, Blott, BH and Daniell, GJ
Optimal Finite Element Modelling and Efficient Reconstruction in Non-Linear 3D Electrical Resistance Tomography.
Proceedings of the 3rd EPSRC Engineering Network Meeting on Biomedical Applications of EIT
University College London, Dr. David Holder, .
Electrical Impedance Tomography can provide images with well-defined characteristics using a fully non-linear reconstruction process when appropriate constraints are imposed on the solution to allow the ill-posed inverse problem to be solved. Using appropriate finite element discretizations for forward solution and inverse problem offers additional advantages in the image reconstruction process, such as (a) inclusion of prior knowledge, (b) generic model templating to adapt to, for example, individual head shapes, and (c) obtaining accurate results without unnecessary computational overhead. We have developed an efficient 3D non-linear reconstruction algorithm based on a regularized inverse conjugate gradient solver which incorporates (a) local image smoothness constraints, and (b) a number of optimisations which reduce the computing power required to obtain an accurate solution. We show results from applying this to various problems which arise in medical resistivity reconstruction given only surface potential measurements and demonstrate the importance of the FE discretization. Keywords: 3D non-linear electrical impedance tomography, FE template modelling, optimal finite element meshes, 3D visualization, FE discretization.
Actions (login required)