Performance comparison study of electrical impedance tomography (EIT) image reconstruction techniques using signal quality metrics and real hardware data

Abstract Electrical impedance tomography (EIT) is a non-invasive imaging technique that reconstructs interior conductivity from surface voltage measurements. This study benchmarks five finite element-based reconstruction algorithms-back projection, Gauss–Newton, Tikhonov regularization, Graz Consensus Reconstruction Algorithm for EIT (GREIT), and Electrical Impedance and Diffuse Optical Reconstruction Software (EIDORS) FEM-under controlled hardware conditions. A 16-electrode saline phantom integrated with multiplexers, an amplifier, a function generator, an Arduino Mega, and a DAQ was used, with chicken breast tissue inserted at four unique positions to mimic tumors. Across four excitation signals (10 kHz/1 Vpp, 50 kHz/1 Vpp, 10 kHz/5 Vpp, 50 kHz/5 Vpp), the methods yielded 80 reconstructed images. Performance was evaluated using quantitative metrics including contrast-to-noise ratio, peak signal-to-noise ratio, structural similarity index, root mean square error, frequency-domain fast Fourier transform, spectrograms, wavelet transform, and spectral temporal energy ratio to evaluate each method’s performance and robustness. According to our investigation, there are notable variations in image fidelity based on the reconstruction technique and the signal circumstances. Back projection exhibited artifacts and lower spatial resolution, whereas GREIT and Tikhonov showed the strongest resilience and noise suppression, with EIDORS offering advantages following them. By integrating real hardware signals with phantom experiments, this work offers a practical benchmarking framework and guidance to enhance medical and industrial EIT systems.