Electrical Impedance Tomography for Thorax Image Using Connected Component Labeling and Sparse Group Lasso

The solution of the inverse problem is essential in electrical impedance tomography. This paper presents a Portable Adjustable Electrical Impedance Tomography System (PAEIT), which consists of 32 electrodes, is a small size, and the excitation frequency, scanning mode and frame rate can be adjusted by keystrokes. Due to the nonlinearity and ill-posedness of electrical impedance tomography, the spatial resolution of the reconstructed image is seriously affected. Therefore, we propose a thorax reconstruction algorithm based on Connected Component Labeling and Sparse Group Lasso (CCL-SGL), which includes CCL-SGL base model (CCL-SGLBM) and CCL-SGL robust model (CCL-SGLRM). This method employs the connected domain labelling algorithm to obtain the structural information of the imaging target from the auxiliary images, which is then adopted as the prior information to provide penalties to the sparse group lasso, which can effectively constrain the freedom of inversion solution and enhance the reconstructed image sparsity. In order to verify the effectiveness of the proposed algorithm, we use numerical simulation, experimental test and real lung datasets for reconstruction experiments, and compare the reconstructed images and quantitative indicators of different algorithms. The results show that the proposed CCL-SGL method improves the spatial resolution and quality of thoracic lung reconstruction images, effectively suppresses background artifacts, and has strong robustness against noise. It is expected to become a general method for EIT image reconstruction.