Adaptive Statistical Error Modeling for Electrical Impedance Tomography With Programmable Resistance Network

Due to the advantage of high temporal resolution, cost-effectiveness, and radiation-free, electrical impedance Tomography (EIT) is considered a promising technique owning a large number of potential industrial and biomedical applications. However, the spatial resolution of EIT is still limited and its imaging results are susceptible to noise. To reduce the impact of measurement noise on the quality of EIT imaging, an adaptive statistical error model (ASEM) is proposed. Unlike noisy models trained by comparing a physical model to its digital twin, ASEM is trained by comparing a digital model to its equal resistance network. The programmable resistance network is configured according to the transfer conductivity matrix derived from the digital model and can be connected to the data acquisition system (DAS) as the physical models. Using the programmable resistance network, a large-scale training dataset can be efficiently collected. To evaluate the performance of the proposed method, a series of experiments were performed with a water tank model. Three different image reconstruction algorithms and one absolute imaging algorithm were tested. The proposed ASEM is trained on 12000 data samples collected by the developed programmable resistance network. The results show that for all tested algorithms, the conductivity reconstruction accuracy is significantly improved using ASEM.