Section: New Results
Computational Anatomy
Statistical Model of the Right Ventricle in Tetralogy of Fallot for Prediction of Cardiac Remodelling
Participants : Tommaso Mansi [ Correspondant ] , Stanley Durrleman, Xavier Pennec, Maxime Sermesant, Nicholas Ayache.
This work has been performed within the Health-e-Child consortium, in close collaboration with several people outside from the lab. In particular I. Voigt (Siemens AG, Erlangen, Germany), B. Bernhardt (McGill University, Montreal Neurological Institute, Montreal, Canada), Dr. A. M. Taylor (Great Ormond Street Hospital, London, UK) and Dr Y. Boudjemline (AP-HP Necker-Enfants Malades, Paris, France).
Understanding and modelling cardiac growth and remodelling is crucial to predict the pathophysiology of patients suffering from repaired tetralogy of Fallot and, therefore, to decide the timing for therapy. In this work, we propose a statistical approach to model cardiac growth in these patients and to exhibit anatomical features that are related to the pathology. From a population of 18 patients (mean age 15 ± 3), we created an average model of the right ventricle anatomy at end diastole by using the unbiased currents-based template estimation method proposed in [32] . Deformation modes are computed to exhibit the variabilities in right ventricle anatomy. Variabilities correlated with body surface area (a continuous clinical index highly correlated with age) and with regurgitation severity are selected using ad-hoc statistical analyses. A generative model of right ventricle growth is obtained from the selected deformation modes. Visual inspection of relevant deformation modes showed realistic patterns according to cardiologists involved in the project. Moreover, the generative statistical model of right ventricle growth has been successfully tested on two new patients. Their body surface area could be predicted by using right ventricle shape only.
Brain morphometry for Alzheimer's disease
Participants : Marco Lorenzi [ Correspondant ] , Xavier Pennec, Giovanni Frisoni [ IRCCS Fatebenefratelli Brescia, Italy ] , Nicholas Ayache.
The work started in October 2009 is actually focused on the definition of a workflow for the robust evaluation of Alzheimer's disease brain changes in datasets of longitudinal structural MRIs.
This will help in the early diagnosis of the disease and in the clinical trials to monitor the effect of new drugs, reducing costs and invasiveness of measurements. Finding surrogate markers to evaluate the progression of Alzheimer's disease is of fundamental importance and currently several measures based on specific functional images are available, but they are expensive and require the injection of mulecular markers.
Several possible algorithms for deformations mapping were investigated, with a particular emphasis on the Symmetric Diffeomorphic Demons. In particular, we are investigating improved measurement of local volume changes.
Statistical regularization of DTI registration
Participants : Andrew Sweet [ Correspondant ] , Xavier Pennec.
In 2008, methods were created to perform efficient non-parameteric registration of diffusion tensor (DT) images using an exact finite strain differential [44] . In this work, we have already adapted these methods to perform the same type of registration in the `log-domain' with symmetric demons forces, as was done in [118] for scalar images. Initial results (see Fig. 12 ) suggest that DT registration exhibits the same benefits from these adaptations as those observed for scalar images. Additionally, these should allow us to easily calculate reliable deformation statistics over a population of diffusion tensor images registered to a general anatomical template. The ultimate goal of the project is to incorporate these statistics back into the regularization scheme, so that we can improve performance and generate smoother deformation fields.
|
