Description
Date depot: 8 septembre 2021
Titre: Modeling changes of dynamics with longitudinal data sets
Directeur de thèse:
Stanley DURRLEMAN (ICM)
Directrice de thèse:
Sophie TEZENAS DU MONTCEL (ICM)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Sciences de l’information et sciences du vivant
Resumé: The study of the temporal progression of a biological or natural phenomenon is central to several scientific fields. For instance, it is used for modeling the changes of biomarkers as a disease progresses.
In the literature, mixed-effects models appear as popular methods for the analysis of longitudinal data. They often rely on a reference time-point such as birth to compare rates of changes from this time-point. Generalization of such methods has been proposed recently to compare trajectories without the prior identification of such a reference time-point and allowing more varied shapes of trajectories than linear. This statistical approach allows the estimation of computational models depicting how a series of biological, imaging, and functional parameters changes during the progression of a neurogedenerative disorder. Nevertheless, this approach allows only the monotonous progression of the biomarkers.
The purpose of this thesis is to use such models of disease progression to measure how an interventional therapy might change the natural progression of a disease. To this end, the candidate will explore ways to add perturbation in the dynamical system driving the model estimation. One may derive and evaluate new ways to measure the efficacy of a treatment by how much the intervention changed the predicted progression. In particular, such an approach would alleviate the hypothesis of monotonicity of the trajectories by taking inspiration from the dynamical systems driving epidemic models.
Doctorant.e: Ortholand Juliette