Research On Reconstruction Algorithm In Magnetic Induction Tomography

Magnetic induction tomography (MIT) is a kind of imaging technology, which uses the principle of electromagnetic detection to measure the conductivity distribution. This technology can be used to achieve the functional imaging and show the image in real time. It is characterized by non-contact, fast imaging, easy to carry and at cost lower. MIT has a good application in clinical noninvasive detection especially in the crania-cerebral diagnoses, which is one significant research subject in biomedical engineering at present.In this research, we do the research on improving the quality of image mainly via the image reconstruction analysis, including solving the forward problem and the images reconstruction. In respect of the forward problem, the variational finite element method is used. By field subdividing and the appropriate interpolation function, we transform the solution of nonlinear partial differential equation into the linear equations. The voltage data can be calculated. In respect of the image reconstruction, in order to improve the quality of the image, a method of modifying the iteration Newton-Raphson (NR) algorithm is presented. In iteration NR, weighting matrix and L1-norm regularization are introduced to overcome the drawbacks of large error and poor stability of image. On the other hand, in order to improve the ill-posed problem, within the incomplete-data framework of the expectation maximization (EM) algorithm, the image reconstruction can be converted to the problem of expectation maximization through the likelihood function. In EM, the missing-data is introduced and the measurement data and the sensitivity matrix are compensated to overcome the drawback that the number of measurement voltage is far less than the number of the unknown. In addition to the above two aspects, the image segmentation is also used, which makes the lesion being more flexible and adaptive to the patients’real conditions, providing a theoretical reference for the development of the application of MIT technique in clinical. The results show that: solving the forward problem with the finite element method can provide the measurement voltage data for image reconstruction; the improved iteration NR method and EM algorithm can improve the quality of image from different points; and the proposed image segmentation can make the lesions of reconstruction image being more likely to be identified.