Project : epidaure

Section: New Results

Motion analysis and medical simulation

Parameters estimation of an electrical model of the heart from in vivo electrical measures.

Keywords : heart modeling, electrophysiology, inverse problem, data assimilation, parameter estimation, reaction-diffusion system, electrical conductivity.

Participants : Valérie Moreau-Villéger, Hervé Delingette, Nicholas Ayache, Sylvain Jaume, Maxime Sermesant, Tristan Picart.

This work was partially funded by the scientific direction of INRIA through the Cooperative Research Action ICEMA2 ( http://www-rocq.inria.fr/sosso/icema2/icema2.html). The electrical measures on canine hearts are provided by Laboratory of Cardiac Energetics, National Heart Lung and Blood Institute, National Institute of Health (E. McVeigh). The electrical measures on human hearts are provided by the Cardiac MR Research Group, King's College London, Guy's Hospital (D. Hill, D. Hawkes, and R. Razavi).

We study the problem of estimating the electrical conductivity of cardiac tissue from a set of temporal in vivo recordings of extracellular potentials. The underlying electrical model is the reaction-diffusion model on the action potential proposed by Aliev and Panfilov. The strategy consists in building an error criterion based upon a comparison of depolarization times between the model and the measures. This error criterion is minimized by a global and then a local adjustment of the model parameters. We have demonstrated the feasibility of this approach on simulated and real in vivo measures on the epicardial surface of a canine heart (Figure 18). This work was published in a research report [88]. We are currently working on the extension of this method to estimate the electrical conductivity on the whole myocardium of a human heart.

Figure 18. Left: conductivity map obtained from our adjustment method for in vivo measures on a canine heart. Middle: depolarization times obtained after a simulation with an homogeneous conductivity map (left) and with the conductivity map (middle) compared with the measures (right). Right: error (in milliseconds) on the estimated depolarization times between simulation and measures.

Image guided laparoscopic spine surgeries

Keywords : 3D/2D Registration, Articulated models, Augmented Reality, Camera Calibration.

Participants : Jonathan Boisvert, Xavier Pennec, Nicholas Ayache.

This project is part of a partnership between the Epidaure team, the Montreal's Sainte-Justine hospital and the Polytechnic School of Montreal (Farida Cheriet).

Laparoscopic spine surgery is a relatively new surgical intervention that has been gaining in popularity in the medical community over the years. The advantages of this technique include small incisions, less blood loss, reduced post-operative pain, earlier discharge from the hospital and a faster post-operative rehabilitation. However, the limited visualization offered to the surgeon is a source of difficulties. One of the solutions to this problem is to develop an augmented reality system which will overlay 3D models of anatomical structures (such as vertebrae or the spinal cord) on the laparoscopic images.

Because the spine is a flexible structure, rigid registration of pre-operative images with per-operative images is not suitable. Moreover, the pre-operative images are taken in standing position, thus variation of the spine's curvature between the two sets of images is great. Therefore, we are investigating the use of an articulated model fitted to each image in order to recover the deformation of the spine. Furthermore, we plan to use the articulated model in order to relate the non-rigid motion of fiducial skin markers with the motion of the spine.

Finally, work has been undertaken to update the laparoscope's calibration during the surgery in order the take into account variations of the laparoscope's internal parameters caused by modifications of the zoom and focus.

Figure 19. Example of 3D vertebrae models overlayed on a laparoscopic image.

Tridimensional Analysis of the heart motion for the study of heart failure

Keywords : cardiac imaging, 3D echography, echocardiography, tracking, segmentation, motion analysis.

Participants : Cécile Marboeuf, Hervé Delingette.

This work is performed in close collaboration with the company Philips Medical Systems Research Paris (Olivier Gérard) in the framework of a 3 year research contract.

The objective of this work, that started in September 2004, is to produce a set of software tools that can help clinicians to better understand, diagnose and cure the phenomenon of cardiac asynchrony, one of the type of heart failure. A first step is to use a geometric heart model for the fine analysis of the patient cardiac motion from acquisition of tridimensional echographic images.

Brain Tumor Growth Simulation

Keywords : Tumor, model, glioblastoma, brain, Magnetic Resonance Imaging, growth, model, simulation, finite element, biomechanics, diffusion, infiltration, mass effect, Clinical Target Volume, Gross Tumor Volume.

Participants : Olivier Clatz, Pierre-Yves Bondiau, Hervé Delingette, Maxime Sermesant, Grégoire Malandain, Nicholas Ayache.

This work was performed in close collaboration with the Surgical Planning Laboratory http://splweb.bwh.harvard.edu:8000/ (Simon Warfield).

We propose a new model to simulate the growth of glioblastomas multiforma (GBM), the most aggressive glial tumors. Because the GBM shows a preferential growth in the white fibers and have a distinct invasion speed with respect to the nature of the invaded tissue, we rely on an anatomical atlas to introduce this information into the model. This atlas includes a white fibers diffusion tensor information and the delineation of cerebral structures having a distinct response to the tumor aggression. We use the finite element method (FEM) to simulate both the invasion of the GBM in the brain parenchyma and its mechanical interaction (mass effect) with the invaded structures. The former effect is modeled with either a reaction-diffusion or a Gompertz equation depending on the considered tissue, while the latter is based on a linear elastic brain constitutive equation. In addition, we propose a new coupling equation taking into account the mechanical influence of the tumor cells on the invaded tissues. This tumor growth model is assessed by comparing the in-silico GBM growth (Figure 20) with the real GBM growth observed between two magnetic resonance images (MRIs) of a patient acquired with six months difference. The quality of the results shows the feasibility of modeling the complex behavior of brain tumors and will justify a further validation of this new conceptual approach.

Figure 20. In Silico growth of a brain tumor, combining diffusion and mechanical effects.

This work has been published in an INRIA Research Report [86], and presented at the 7th International Conference on Medical Image Computing and Computer-Assisted Intervention [55].