Study Of Image Reconstruction In Electrical Impedance Tomography Using The Algorithm Based On Maximum A Posteriori

Electrical Impedance Tomography (EIT), a type of electromagnetic imaging modalities, provides a new technique for medical diagnosis. It calculates an estimation of the conductivity distribution or variation within a body based on current stimulations and voltages measurements on the body surface. It is a kind of easy-to-design, convenient, inexpensive, especially invasive and real-time medical device. EIT group of FMMU made breakthroughs in early preliminary research work and developed related clinical experiments on particular patients, especially on those who needed bedside monitoring. In clinical applications, however, many practical factors affected the qualities of EIT imaging. For example, uncertainty caused by subjects'breathing or posture changes brought artifacts to EIT images.In order to address these issues and improve the qualities of EIT imaging relatively, two phases of EIT reconstruction problems were discussed as follows:(1) Comparison of EIT reconstruction algorithmsReconstruction Algorithm is a fundamental and crucial part of EIT. The nature of algorithm is to solver the ill-posedness of EIT inverse problem, and to balance a tradeoff between image resolution and noise performance. Moreover, efficiency of reconstruction and stabilities of algorithms must be under consideration. Based on the theory of Bayes decision and the maximum likelihood estimation method, the study discussed a maximum a posteriori (MAP) approach to reconstruct EIT images. The principle of MAP is to estimate the most possible conductivity based on probabilistic model and measurements. The approach considers variations not only of conductivity but also of measurement noise. The study compared MAP approach to the traditional equipotential backprojection (EPBP) approach through simulations, phantom experiments and lung ventilation. In simulations, MAP improved spatial resolution by average 78% and positioning accuracy by average 70%; in phantom experiments, MAP improved spatial resolution by average 85% and positioning accuracy by average 46%, compared to EPBP. Under the same hardware setup, MAP consumed less than 50ms for online calculation as EPBP did and reconstructed available EIT image constantly. Preliminary results showed that MAP performed better spatial resolution and positioning accuracy than EPBP and consumed approximate reconstructing time. Therefore, MAP is optimal approach on EIT inverse solution in clinical applications.(2) Improvement of EIT algorithm for clinical applicationsIn clinical applications, the high ill-posedness of EIT makes reconstructed images vulnerable to measurement disturbance and noise, which results in terrible artifacts in EIT image. Therefore, it is important to reject disturbing signals and reduce data noise.EIT has excellent temporal resolution, which makes EIT a promising technology to monitor the conductivity changes caused by fast physiological events, and more importantly, reduces EIT image noise. Most EIT reconstruction algorithms solve each data frame independently. However, Kalman Filter algorithm tracks the image changes across frames. It estimates current solution iteratively based on previous solution and current measurements. This study discussed a direct reconstruction approach based on MAP, which also considered 3 previous and future data frames to calculate current images. Simulation results showed that the latter algorithm had a better noise performance even under strong noise background (0.5 noise level).One key difficulty with EIT measurements is due to the contact and position uncertainty of the electrodes, in which the body surface moves during breathing and posture changes. Under these circumstances, the quality of EIT images is very poor. The study developed an extension of the MAP approach which directly reconstructed both internal conductivity changes and electrode movements for difference EIT using an augmented reconstruction matrix. Simulation results showed that standard MAP could not reconstruct the targets in EIT image when 1% electrode movement occurred. However, the MAP extension was able to reconstruct the targets and reduced artifacts dramatically and it also showed the situation of electrode movement to provide valuable information for researchers.