Deep Learning based Metal Artifacts Reduction in post-operative Cochlear Implant CT Imaging - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Communication Dans Un Congrès Année : 2019

Deep Learning based Metal Artifacts Reduction in post-operative Cochlear Implant CT Imaging

Résumé

To assess the quality of insertion of Cochlear Implants (CI) after surgery, it is important to analyze the positions of the electrodes with respect to the cochlea based on post-operative CT imaging. Yet, these images suffer from metal artifacts which often entail a difficulty to make any analysis. In this work, we propose a 3D metal artifact reduction method using convolutional neural networks for post-operative cochlear implant imaging. Our approach is based on a 3D generative ad-versarial network (MARGANs) to create an image with a reduction of metal artifacts. The generative model is trained on a large number of pre-operative "artifact-free" images on which simulated metal artifacts are created. This simulation involves the segmentation of the scala tym-pani, the virtual insertion of electrode arrays and the simulation of beam hardening based on the Beer-Lambert law. Quantitative and qualitative evaluations compared with two classical metallic artifact reduction algorithms show the effectiveness of our method.
Fichier principal
Vignette du fichier
paper2371.pdf (631.62 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-02196557 , version 1 (29-07-2019)

Identifiants

  • HAL Id : hal-02196557 , version 1

Citer

Zihao Wang, Clair Vandersteen, Thomas Demarcy, Dan Gnansia, Charles Raffaelli, et al.. Deep Learning based Metal Artifacts Reduction in post-operative Cochlear Implant CT Imaging. MICCAI 2019 - 22nd International Conference on Medical Image Computing and Computer Assisted Intervention, Oct 2019, Shenzhen, China. pp.121-129. ⟨hal-02196557⟩
280 Consultations
641 Téléchargements

Partager

Gmail Facebook X LinkedIn More