Impact Study and Evaluation of SNR on Different Electrical Impedance Tomography Reconstruction Algorithms

This work explored the application effect of different electrical impedance tomography(EIT) image reconstruction algorithms in the imaging of single target with different positions or multi-targets when using detection system with different signal-to-noise ratio(SNR).Aiming at EIT system with 16 electrodes,ideal boundary voltage of EIT field obtained by simulation was taken as signal acquired by ideal EIT system,gauss white noise was added to the whole data collected by all the channels of ideal system,so as to simulate the measuring voltage collected by EIT detection systems of which SNR are 40 dB,60 dB and 80 dB.The linear back-projection algorithm(LBP),Tikhonov regularization,the combined Tikhonov-Noser regularization algorithm,the Landweber iterative and Newton-Raphson algorithm were adopted in the image reconstruction.Imaging error function ER and structure similarity degree SSIM were introduced to evaluate the reconstructed images.With the movement of a single target from the center to the edge of the field,the value of SSIM of all the 5 algorithms increased.For the targets near to the field edge,the value of ER of all the 5 algorithms increased with the increase of targets number.Both LBP and the combined Tikhonov-Noser regularization algorithm successfully produced images for all the 5 targets supposed in this study under the three EIT systems with different SNR.Of all the 5 algorithms,Tikhonov-Noser regularization algorithm obtained the best image reconstruction quality.For the target A near to the field centre,Newton-Raphson algorithm was unable to image when the SNR was 40 dB,while the SNR which Tikhonov and Landweber algorithms are unable to image was from 40 dB to 60 dB.When SNR was 80 dB,image reconstruction quality of LBP was the lowest of all the 5 algorithms.Different algorithms had different requirements to SNR of EIT detection system,the index SNR should be designed according to concrete application target and algorithms adopted when EIT detection system is constructed.Generally,the combined Tikhonov-Noser regularization algorithm was preferred.When the SNR of the system under 60 dB,LBP was advised and it should be abandoned when SNR was over 80 dB.Newton-Raphson is advised when the SNR was about 60 dB.