Application of neural networks inversion in Magnetopneumography

Issue title: OIPE 2006

Article type: Research Article

Authors: Tomek, Jiří^{a; *} | Platil, Antonín^a | Ripka, Pavel^a

Affiliations: [a] Czech Technical University, Faculty of Electrical Engineering, Department of Measurement, Technická 2,166 27 Prague, Czech Republic

Correspondence: [*] Corresponding author. Tel.: +420 2 2435 2178; Fax: +420 2 3333 9929; E-mail: tomekj1@fel.cvut.cz

Abstract: To obtain unknown source distribution and magnetic moments of discrete magnetized ferromagnetic elementary volumes in our lung phantom we have chosen neural network inverse model. The simplified task was solved by regression multiple-layer perceptron network taught on data generated by multi-pole discrete forward model of the lungs. Magnetic fields corresponding to numerous possible sources were processed. The results are quite promising though more complex cases and especially in-vivo measurements would require much more measuring nodes over the patient/phantom to get enough data for reasonable inversion.

Keywords: Magnetopneumography, inversion, modeling, neural networks

DOI: 10.3233/JAE-2007-919

Journal: International Journal of Applied Electromagnetics and Mechanics, vol. 26, no. 3-4, pp. 285-290, 2007