Research On Image Reconstruction Algorithms For Biomedical Electrical Impedance Tomography

Electrical Impedance Tomography (EIT) imaging technology is developed over 30 years with advantages of non-invasive, portability, low cost, rapid response compared to the other imaging technologies. It has potential application background in the industrial and medical fields. It is a new generation of medical imaging technology in recent years after morphology and structural imaging. Due to EIT generating and measuring safe electrical signals, the human body avoids exposuring under ray and nuclide which makes it becomes a hotspot research field as an ideal and a promising noninvasive medical imaging technology.The key problem of EIT is how to solve inverse problem. Due to the non-linearlity and ill-posed features of the soft-field, different reconstruction algorithm will get different solutions, and also reconstructs different images. In this thesis, EIT reconstruction algorithms are researched in detail.This thesis is organised and covered the following aspects:1. As the basement of reconstruction, EIT forword models are establised up by COMSOL Multiphysics.16 electrodes attached circling the boundary uniformly and non-uniformly are discussed to verify the effection of electrodes placement to the sensitivity distribution within the field. Three working patterns, adjacent, opposite and interval current injections are studied to get optimal working pattern.2. Based on the solution of the forward problem of the optimal working pattern, Jacobian matrix is achieved. Reconstruction algorithms are presented in the thesis including SDA (Steepest Descent Algorithm) iteration algorithm, Newton-Raphson iteration algorithm, MNR (Modified Newton-Raphson) iteration algorithm, NOSER (Newton’s one-step error reconstruction) algorithm, the standard Tikhonov regularization algorithm and a Total Variation regularization algorithm.3. The qualities of the reconstructed images by the above image reconstruction algorithms are evaluated with correlation coefficient and relative error indexes.4. Besides the circle model, an octagon field model is studied hoping to improve the space resolution. Through comparison the uniformity of the circle and octagon fields’ sensitivity matrix, it is verified that the octagon field can get higher spatial resolution, which may offer a theoretical foundation for EIT to put forward into more applications.