Acceleration Of Electrical Impedance Tomography Algorithm With Open And Closed Domain Model Evaluation

Abstract

The presented paper discusses the possibilities of parallelizing algorithms for the image reconstruction in the Electrical Impedance Tomography (EIT). The parallelization process was performed and tested with combination of Matlab and CUDA environment. A twenty-fold acceleration of the Jacobian calculation was achieved. Besides of the parallelization process, the important part of the EIT experiments was the selection of the computational mesh. These mesh elements are used for fast execution of the partial differential equation EIT model. Forward and inverse problems of the image reconstruction were simulated on the novel open domain. The optimized domain reduced error in the reconstructed image.