EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
2518. Real data EIT reconstruction using virtual X-rays and deep learning
Invited abstract in session MC-34: Optimization and learning for data science and imaging (Part III), stream Advances in large scale nonlinear optimization.
Monday, 12:30-14:00Room: 43 (building: 303A)
Authors (first author is the speaker)
| 1. | Siiri Rautio
|
| University of Helsinki |
Abstract
In electrical impedance tomography (EIT), the aim is to recover the unknown conductivity of a target by injecting currents and measuring boundary voltages through electrodes. It is a nonlinear and severely ill-posed inverse problem. We introduce a new reconstruction algorithm for EIT, which provides a connection between EIT and traditional X-ray tomography, based on the idea of "virtual X-rays". We divide the exponentially ill-posed and nonlinear inverse problem of EIT into separate steps. We start by mathematically calculating so-called virtual X-ray projection data from the measurement data. Then we perform explicit algebraic operations and one-dimensional integration, ending up with a blurry and nonlinearly transformed Radon sinogram. We use neural networks to remove the higher-order scattering terms and perform deconvolution. Finally, we can compute a reconstruction of the conductivity using the inverse Radon transform. We demonstrate the method with experimental data. This is a joint work with Melody Alsaker, Fernando Moura, Juan Pablo Agnelli, Rashmi Murthy, Matti Lassas, Jennifer Mueller, and Samuli Siltanen.
Keywords
- Algorithms
- Machine Learning
- Computer Science/Applications
Status: accepted
Back to the list of papers