Segmentation and registration coupling from short-axis Cine MRI: application to infarct diagnosis

Abstract : Estimating regional deformation of the myocardium from Cine MRI has the potential to locate abnormal tissue. Regional deformation of the left ventricle is commonly estimated using either segmentation or 3D+t registration. Segmentation is often performed at each instant separately from the others. It can be tedious and does not guarantee temporal causality. On the other hand, extracting regional parameters through image registration is highly dependent on the initial segmenta-tion chosen to propagate the deformation fields and may not be consistent with the myocardial contours. In this paper, we propose an intermediate approach that couples segmentation and registration in order to improve temporal causality while removing the influence of the chosen initial segmentation. We propose to apply the deformation fields from image registration (sparse Bayesian registration) to every segmentation of the cardiac cycle and combine them for more robust regional measurements. As an illustration, we describe local deformation through the measurement of AHA regional volumes. Maximum regional volume change is extracted and compared across scar and non-scar regions defined from delayed enhancement MRI on 20 ST-elevation myocardial infarction patients. The proposed approach shows (i) more robustness in extracting regional volumes than direct segmentation or standard registration and (ii) better performance in detecting scar.
Type de document :
Communication dans un congrès
7th International Statistical Atlases and Computational Modeling of the Heart (STACOM) Workshop, Held in Conjunction with MICCAI 2016, Oct 2016, Athens, Greece. 10124, pp.48-56, 2016, Lecture Notes in Computer Science
Liste complète des métadonnées

Littérature citée [13 références]  Voir  Masquer  Télécharger

https://hal.inria.fr/hal-01352460
Contributeur : Nicolas Duchateau <>
Soumis le : lundi 8 août 2016 - 10:11:20
Dernière modification le : lundi 9 avril 2018 - 09:17:09
Document(s) archivé(s) le : mercredi 9 novembre 2016 - 12:17:31

Fichier

cMarchesseau_STACOM_2016.pdf
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-01352460, version 1

Collections

Citation

Stephanie Marchesseau, Nicolas Duchateau, Hervé Delingette. Segmentation and registration coupling from short-axis Cine MRI: application to infarct diagnosis. 7th International Statistical Atlases and Computational Modeling of the Heart (STACOM) Workshop, Held in Conjunction with MICCAI 2016, Oct 2016, Athens, Greece. 10124, pp.48-56, 2016, Lecture Notes in Computer Science. 〈hal-01352460〉

Partager

Métriques

Consultations de la notice

205

Téléchargements de fichiers

189