Robust Electrical Impedance Tomography via Half-Quadratic Optimization

Electrical impedance tomography (EIT) is a promising imaging technique in the medical field. However, its clinical application is limited by several challenges, such as poor electrode contact and body movement. These factors introduce impulsive noise in boundary voltage measurement, resulting in the performance degradation of the conventional imaging methods. This letter devises a robust approach to improve the performance in such scenarios. Unlike traditional algorithms that use an identity matrix as the measurement noise weight matrix, the suggested method designs a strategy to update the weight matrix during iteration for outlier resistance. Specifically, we apply the penalty technique to optimize the weight matrix and thus formulate the imaging task as a half-quadratic optimization problem. In addition, we exploit a median absolute deviation based confidence interval to distinguish outlier-contaminated and normal channels, such that the elements of the diagonal weight matrix are either 0 or 1. Subsequently, the resultant optimization task is addressed using the alternating minimization method. Experimental results show that the proposed approach outperforms the state-of-the-art algorithms in the presence of impulsive noise.