Ultrasound Tomography Imaging of Defects Using Neural Networks
Abstract
Simulations of ultrasound tomography demonstrated that artificial neural networks can solve the inverse problem in ultrasound tomography. A highly simplified model of ultrasound propagation was constructed, taking no account of refraction or diffraction, and using only longitudinal wave time of flight (TOF). TOF data were used as the network inputs, and the target outputs were the expected pixel maps, showing defects (gray scale coded) according to the velocity of the wave in the defect. The effects of varying resolution and defect velocity were explored. It was found that defects could be imaged using time of flight of ultrasonic rays.
Cite
Text
Anthony et al. "Ultrasound Tomography Imaging of Defects Using Neural Networks." Neural Computation, 1992. doi:10.1162/NECO.1992.4.5.758Markdown
[Anthony et al. "Ultrasound Tomography Imaging of Defects Using Neural Networks." Neural Computation, 1992.](https://mlanthology.org/neco/1992/anthony1992neco-ultrasound/) doi:10.1162/NECO.1992.4.5.758BibTeX
@article{anthony1992neco-ultrasound,
title = {{Ultrasound Tomography Imaging of Defects Using Neural Networks}},
author = {Anthony, Denis M. and Hines, Evor L. and Hutchins, David A. and Mottram, J. Toby},
journal = {Neural Computation},
year = {1992},
pages = {758-771},
doi = {10.1162/NECO.1992.4.5.758},
volume = {4},
url = {https://mlanthology.org/neco/1992/anthony1992neco-ultrasound/}
}