A neural network based method for the inverse problem of electrical impedance tomography

Electrical impedance tomography is an emerging imaging modality based on reconstruction of the resistivity distribution of the interior of the body, using non-invasive surface electrical measurements. Image reconstruction is a nonlinear, ill-posed, inverse problem. A new approach to the reconstruction problem is proposed, using multi-layer feed forward neural networks. The technique was tested using numerical simulations and actual measurements from a physical model. When compared with iterative techniques, superior performance was observed vis-a-vis image quality, reconstruction time and robustness to noise. This method can be easily adapted for different forward models of resistivity distribution used in existing reconstruction algorithms.