RESIST: Remapping EIT Signals Using Implicit Spatially-Aware Transformer

Dominik Becker, Anita Just, Günter Hahn, Peter Herrmann, Leif Saager, Fabian H. Sinz
Proceedings of the 4th Machine Learning for Health Symposium, PMLR 259:86-103, 2025.

Abstract

Electrical impedance tomography (EIT) aims to reconstruct the body’s internal electrical conductivity distribution from surface voltage measurements. This non-invasive, non-ionizing, and cost-effective technique is valuable for medical applications, offering potential for long- term monitoring of lung functionality and condition. However, existing methods of absolute EIT provide blurred tomograms that are difficult to interpret, do not resemble the human topography and are therefore of limited use for clinical applications. We propose RESIST, a new data-driven approach that integrates prior geometry of human bodies and conductivity information by combining an implicit neural network with a transformer model in an encoder-decoder framework. RESIST maps simultaneous EIT measurements from multiple body levels to conductivity values at any coordinate in a 3D body volume. We train RESIST on simulated EIT measurements based on 3D human body CT segmentations. We find that it is robust against distortions in the signal and exact placement of electrodes, correctly infers interpolation laws, and generalizes to real EIT measurements in humans.

Cite this Paper

BibTeX
@InProceedings{pmlr-v259-becker25a, title = {RESIST: Remapping EIT Signals Using Implicit Spatially-Aware Transformer}, author = {Becker, Dominik and Just, Anita and Hahn, G{\"{u}}nter and Herrmann, Peter and Saager, Leif and Sinz, Fabian H.}, booktitle = {Proceedings of the 4th Machine Learning for Health Symposium}, pages = {86--103}, year = {2025}, editor = {Hegselmann, Stefan and Zhou, Helen and Healey, Elizabeth and Chang, Trenton and Ellington, Caleb and Mhasawade, Vishwali and Tonekaboni, Sana and Argaw, Peniel and Zhang, Haoran}, volume = {259}, series = {Proceedings of Machine Learning Research}, month = {15--16 Dec}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v259/main/assets/becker25a/becker25a.pdf}, url = {https://proceedings.mlr.press/v259/becker25a.html}, abstract = {Electrical impedance tomography (EIT) aims to reconstruct the body’s internal electrical conductivity distribution from surface voltage measurements. This non-invasive, non-ionizing, and cost-effective technique is valuable for medical applications, offering potential for long- term monitoring of lung functionality and condition. However, existing methods of absolute EIT provide blurred tomograms that are difficult to interpret, do not resemble the human topography and are therefore of limited use for clinical applications. We propose RESIST, a new data-driven approach that integrates prior geometry of human bodies and conductivity information by combining an implicit neural network with a transformer model in an encoder-decoder framework. RESIST maps simultaneous EIT measurements from multiple body levels to conductivity values at any coordinate in a 3D body volume. We train RESIST on simulated EIT measurements based on 3D human body CT segmentations. We find that it is robust against distortions in the signal and exact placement of electrodes, correctly infers interpolation laws, and generalizes to real EIT measurements in humans.} }
Endnote
%0 Conference Paper %T RESIST: Remapping EIT Signals Using Implicit Spatially-Aware Transformer %A Dominik Becker %A Anita Just %A Günter Hahn %A Peter Herrmann %A Leif Saager %A Fabian H. Sinz %B Proceedings of the 4th Machine Learning for Health Symposium %C Proceedings of Machine Learning Research %D 2025 %E Stefan Hegselmann %E Helen Zhou %E Elizabeth Healey %E Trenton Chang %E Caleb Ellington %E Vishwali Mhasawade %E Sana Tonekaboni %E Peniel Argaw %E Haoran Zhang %F pmlr-v259-becker25a %I PMLR %P 86--103 %U https://proceedings.mlr.press/v259/becker25a.html %V 259 %X Electrical impedance tomography (EIT) aims to reconstruct the body’s internal electrical conductivity distribution from surface voltage measurements. This non-invasive, non-ionizing, and cost-effective technique is valuable for medical applications, offering potential for long- term monitoring of lung functionality and condition. However, existing methods of absolute EIT provide blurred tomograms that are difficult to interpret, do not resemble the human topography and are therefore of limited use for clinical applications. We propose RESIST, a new data-driven approach that integrates prior geometry of human bodies and conductivity information by combining an implicit neural network with a transformer model in an encoder-decoder framework. RESIST maps simultaneous EIT measurements from multiple body levels to conductivity values at any coordinate in a 3D body volume. We train RESIST on simulated EIT measurements based on 3D human body CT segmentations. We find that it is robust against distortions in the signal and exact placement of electrodes, correctly infers interpolation laws, and generalizes to real EIT measurements in humans.
APA
Becker, D., Just, A., Hahn, G., Herrmann, P., Saager, L. & Sinz, F.H.. (2025). RESIST: Remapping EIT Signals Using Implicit Spatially-Aware Transformer. Proceedings of the 4th Machine Learning for Health Symposium, in Proceedings of Machine Learning Research 259:86-103 Available from https://proceedings.mlr.press/v259/becker25a.html.

Related Material