Skip to Main content Skip to Navigation

Image-based physiological and statistical models of the heart: application to tetralogy of Fallot

Tommaso Mansi 1
1 ASCLEPIOS - Analysis and Simulation of Biomedical Images
CRISAM - Inria Sophia Antipolis - Méditerranée
Abstract : This thesis presents image-based methods for diagnosis, prognosis and therapy planning of patients with repaired tetralogy of Fallot, a severe congenital heart disease. This was achieved by combining advanced medical image processing with both statistical and physiological modelling. First, we proposed a demons-based image registration algorithm to estimate the 3D myocardium strain from routine medical images, which is challenging due to the aperture problem (only the apparent motion is visible). The algorithm relies on elasticity and incompressibility constraints, rigorously implemented thanks to a new justification of the demons regularisation. Experiments on synthetic data and cine MRI of patients demonstrated that the proposed constraints improve the accuracy of the estimated cardiac displacements. Second, a generative model of the pathological right ventricle growth was proposed using a well-posed statistical framework for shape analysis, based on the "currents" shape representation. Principal component analysis was used to identify shape features that are relevant to the disease. Partial least squares and canonical correlation analysis were employed to design a generative model of heart growth. Applied on the right ventricles of 32 patients with tetralogy of Fallot, the method identified the right ventricle dilation, basal bulging and apical dilation reported in the clinical literature. The model showed that these features appear progressively as the child grows. Finally, we introduced an electromechanical model of the heart for personalised simulations of valve replacement in repaired tetralogy of Fallot patients. The electromechanical model simulates the main features of the cardiac function observed in these patients. Once personalised from clinical data, the model was used to predict the effects of pulmonary valve replacement on the postoperative cardiac function. Tested on two patients, the model managed to qualitatively reproduce their cardiac function. As expected, valve replacement predicted a significant improvement of the right ventricle but also of the left ventricle, suggesting a tight inter-ventricular relationship. The combination of medical imaging, statistical analysis and biophysical modelling provides a powerful framework for a more personalised computer-aided medicine.
Document type :
Complete list of metadatas
Contributor : Tommaso Mansi <>
Submitted on : Sunday, October 31, 2010 - 3:28:51 PM
Last modification on : Tuesday, October 20, 2020 - 10:20:17 AM
Long-term archiving on: : Friday, October 26, 2012 - 12:50:11 PM


  • HAL Id : tel-00530956, version 1



Tommaso Mansi. Image-based physiological and statistical models of the heart: application to tetralogy of Fallot. Human-Computer Interaction [cs.HC]. École Nationale Supérieure des Mines de Paris, 2010. English. ⟨NNT : 2010ENMP0023⟩. ⟨tel-00530956⟩



Record views


Files downloads