Standardization of Multicentric Image Datasets with Generative Adversarial Networks - Archive ouverte HAL
Communication Dans Un Congrès Année : 2019

Standardization of Multicentric Image Datasets with Generative Adversarial Networks

Résumé

Due to the sensitivity of medical images to acquisition parameters, multicentric image studies often suffer from a lack of homogeneity in terms of statistical characteristics, also known as the center effect. There is therefore a clear need for image-based standardization techniques. In this paper, we propose a two-step machine learning based framework in which multicentric, heterogeneous images are translated to match the statistical properties of a standard domain. We apply our standardization model to a publicly available multicentric dataset, where we show that we reduce cross-domain while preserving within-domain variability.
Fichier principal
Vignette du fichier
Hognon_MIC2019.pdf (278.41 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-02447807 , version 1 (21-01-2020)

Identifiants

  • HAL Id : hal-02447807 , version 1

Citer

Clément Hognon, Florent Tixier, Olivier Gallinato, Thierry Colin, Dimitris Visvikis, et al.. Standardization of Multicentric Image Datasets with Generative Adversarial Networks. IEEE Nuclear Science Symposium and Medical Imaging Conference 2019, Oct 2019, Manchester, United Kingdom. ⟨hal-02447807⟩

Collections

UNIV-BREST LATIM
407 Consultations
394 Téléchargements

Partager

More