Abstract

The paper proposes a GAN-based Metal Artifact Reduction (MAR) method that relies on simulated training data and is suitable for pre- and postoperative images. This approach is the first MAR algorithm that combines the physical simulation of metal artifacts with 3D GAN networks. The method is based on two main stages: the simulation of the presence of metal parts in the images and the simulation of the creation of artifacts caused by those metal parts. The MARGAN method was applied to a set of inner ear CT images to reduce the artifacts created by cochlear implants.

Key Words: Metal Artifact Reduction, GAN, CT images, cochlear implants, artifact simulation

Main points

Proposes a GAN-based Metal Artifact Reduction (MAR) method that relies on simulated training data.
Introduces the first MAR algorithm that combines the physical simulation of metal artifacts with 3D GAN networks.
Demonstrates the application of the MARGAN method to a set of inner ear CT images to reduce the artifacts created by cochlear implants.

Citation

@article{WANG2021101990,
title = {Inner-ear augmented metal artifact reduction with simulation-based 3D generative adversarial networks},
journal = {Computerized Medical Imaging and Graphics},
volume = {93},
pages = {101990},
year = {2021},
issn = {0895-6111},
doi = {https://doi.org/10.1016/j.compmedimag.2021.101990},
url = {https://www.sciencedirect.com/science/article/pii/S0895611121001397},
author = {Zihao Wang and Clair Vandersteen and Thomas Demarcy and Dan Gnansia and Charles Raffaelli and Nicolas Guevara and Hervé Delingette},
keywords = {Artifact reduction, Deep learning, GAN},
abstract = {Metal Artifacts creates often difficulties for a high quality visual assessment of post-operative imaging in computed tomography (CT). A vast body of methods have been proposed to tackle this issue, but these methods were designed for regular CT scans and their performance is usually insufficient when imaging tiny implants. In the context of post-operative high-resolution CT imaging, we propose a 3D metal artifact reduction algorithm based on a generative adversarial neural network. It is based on the simulation of physically realistic CT metal artifacts created by cochlea implant electrodes on preoperative images. The generated images serve to train a 3D generative adversarial networks for artifacts reduction. The proposed approach was assessed qualitatively and quantitatively on clinical conventional and cone beam CT of cochlear implant postoperative images. These experiments show that the proposed method outperforms other general metal artifact reduction approaches.}
}