Electrical impedance tomography image reconstruction based on backprojection and extreme learning machines

• Published in: Research on biomedical engineering Vol. 36; no. 4; pp. 399 - 410
• Main Authors: Gomes, Juliana C., Barbosa, Valter A. F., Ribeiro, David E., de Souza, Ricardo E., dos Santos, Wellington Pinheiro
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.12.2020
• Subjects: Biomaterials; Biomedical Engineering and Bioengineering; Biomedical Engineering/Biotechnology; Engineering; Original Article; Electrical impedance tomography; Extreme learning machines; Modified backprojection; Image reconstruction; Artificial neural networks
• ISSN: 2446-4732; 2446-4740
• DOI: 10.1007/s42600-020-00079-3

Purpose Electrical impedance tomography (EIT) is an image technique based on the application of an alternating electrical current on electrodes placed on the surface of the domain, which are also responsible for measuring the resulting electrical potentials. EIT main advantages are portability, low cost, and nonuse of ionizing radiation. EIT image reconstruction depends on the resolution of the direct and inverse problems, which is nonlinear and ill-posed. Several reconstruction methods have been used to solve EIT inverse problem, from Newton-based methods to bio- and social-inspired metaheuristics. Method In this work, we propose a new approach: the use of random-weighted neural networks, specifically extreme learning machines (ELM), to approximate sinograms from electrical potential data and, therefore, use the classical backprojection algorithm for image reconstruction. We generated a database of 4000 synthetic 128 × 128 images. We trained 16 ELMs corresponding to a 16-electrode EIT system placed on a circular domain. Results Results were evaluated according to peak-to-noise ratio (PSNR) and Structural Similarity Index (SSIM), as well as visual inspection. Results are fair and similar to image reconstructions obtained by the direct application of the backprojection algorithm, in case this reconstruction problem would be a classic tomographic reconstruction. Conclusion Our work suggested the use of ELMs to create sinograms from the electrical potential data of the studied domain. Sinograms could be easily reconstructed by the backprojection technique for final image generation, where objects and domain are represented. Finally, we were able to reduce the time of EIT reconstruction, mainly due to the high speed of the ELMs, and obtain reasonable and consistent images.