Archive ouverte HAL - Personalization of Synthetic Pathological Left Ventricles using Learning-Based Infarct Localization and Finite-Element Inverse Modeling

Personalization of Synthetic Pathological Left Ventricles using Learning-Based Infarct Localization and Finite-Element Inverse Modeling

G Kenny Rumindo 1 Nicolas Duchateau 1 Jacques Ohayon 2 Pierre Croisille 3, 4 Patrick Clarysse 1
1 MOTIVATE - Imagerie et modélisation Vasculaires, Thoraciques et Cérébrales
CREATIS - Centre de Recherche en Acquisition et Traitement de l'Image pour la Santé : 139739
2 TIMC-IMAG-DyCTiM - Dynamique Cellulaire et Tissulaire- Interdisciplinarité, Modèles & Microscopies
TIMC-IMAG - Techniques de l'Ingénierie Médicale et de la Complexité - Informatique, Mathématiques et Applications, Grenoble - UMR 5525
3 RMN et optique : De la mesure au biomarqueur
CREATIS - Centre de Recherche en Acquisition et Traitement de l'Image pour la Santé
Abstract : Introduction Cardiac ischemia is a condition in which the coronary arteries are narrowed, restricting the blood flow to the heart muscles and causing myocardial infarction (MI). The healing process after MI is called myocardial remodeling, and its prognosis is essential to predict the heart condition [1]. However, the mechanisms involved in myocardial remodeling are still not well understood. Our aim is to gain more understanding of the remodeling process by obtaining personalized indices that relate more to the mechanical functionality of the heart than tissue viability or strain values. Method We introduced a personalization approach to better assess the regional myocardial functional status. The approach combines state-of-the-art infarct localization [2] and a novel finite-element (FE) cardiac modeling [3,4]. Based on the personalized model, we propose stiffness-related indices that would provide more information on the heart condition. Fig.1-top describes the proposed personalization approach. From cardiac magnetic resonance (CMR) data, we reconstruct the left ventricle (LV) geometry, extract the strain data, the LV volume and pressure if available. From the strain pattern, abnormal regions are estimated via a learning-based method to partition the LV geometry into healthy and potentially-infarcted regions. Next, we incorporate a rule-based fiber orientation and implement a constitutive law to define the mechanical behavior of the LV [4]. Optimization is finally performed to personalize the material parameters for each case. The objective function is based on the strain data, volume and pressure. Once satisfied, the personalized material parameters are available for each region. Results This study verified the feasibility of the approach by applying it to 5 pathological cases obtained from forward modeling of actual LV geometries. Fig.1-bottom compares the ground truth and the results for one case: the infarct prediction (left) and the strain pattern distribution (middle). Fig.1-bottom (right) lists the ground truth and the personalized parameters of the same case. This finding confirms that the suggested approach was able to personalize each case within an acceptable range of accuracy. Conclusion We investigated the feasibility of combining infarct localization and cardiac modeling to personalize pathological LV for better assessment of heart conditions post-MI. Future studies built on this approach on clinical data are necessary, primarily to investigate the correlation between the personalized material parameters and the real condition of the patient through several time-points of CMR acquisitions, in order to predict the future condition of the patient’s heart. Acknowledgements This project has received funding from the European Commission’s Horizon2020 Marie Sklodowska-Curie ETN VPH-CaSE (www.vph-case.eu), agreement No. 642612 References [1] Flaschskampf et al. Eur Heart J: 32(3), 272-283 [2] Rumindo et al. FIMH 2017: 106-114 [3] Rumindo et al. CMBBE 2016: 161-169 [4] Genet et al. J Appl Physiol (2014): 117, 142-152
Liste complète des métadonnées

Littérature citée [4 références]  Voir  Masquer  Télécharger

https://hal.archives-ouvertes.fr/hal-01912922
Contributeur : Béatrice Rayet <>
Soumis le : mercredi 20 mars 2019 - 16:54:50
Dernière modification le : mardi 26 mars 2019 - 10:10:56
Document(s) archivé(s) le : vendredi 21 juin 2019 - 22:04:12

Fichier

Rumindo_AbstractSubmission_WCB...
Fichiers produits par l'(les) auteur(s)

Identifiants

  • HAL Id : hal-01912922, version 1

Citation

G Kenny Rumindo, Nicolas Duchateau, Jacques Ohayon, Pierre Croisille, Patrick Clarysse. Personalization of Synthetic Pathological Left Ventricles using Learning-Based Infarct Localization and Finite-Element Inverse Modeling. 2018 World Congress of Biomechanics, Jul 2018, Dublin, Ireland. ⟨hal-01912922⟩

Partager

Métriques

Consultations de la notice

85

Téléchargements de fichiers

7