Influence of hyperparameter on the redistribution index in an electrical impedance tomography algorithm Using discrete cosine transformation

Abstract

With the introduction of a structural prior, the images of electrical impedance tomography (EIT) benefit from the improvement of the interpretability in clinical settings. However, the improvement comes with the risk of the outdated structural prior which does not comply with the current patient status, therefore resulting in a misleading reconstruction then compromising the clinical decision. The redistribution index can detect an outdated structural prior by quantitatively analyzing EIT reconstructions. The choice of the hyperparameter ? in DCT-based EIT algorithm influences the EIT reconstructions in addition to the structural priors. In this contribution, the influence of hyperparameters on redistribution index was investigated by means of simulations. We conducted a series of simulations in terms of 26  different scales of dorsal lung atelectasis, then the simulation data were reconstructed with 20 different hyperparameters, at last the EIT reconstructions were used to investigate the behavior of redistribution index. The result reveals that the function of redistribution index to detect an outdated structural prior is rather robust regardless of the optimal hyperparameter.