ED Mathématiques et Informatique
STUDY AND DESIGN OF INTERACTIVE IMMERSIVE VISUALIZATION EXPERIENCES FOR NON EXPERTS
by Edwige GROS (LaBRI - Laboratoire Bordelais de Recherche en Informatique)
The defense will take place at 14h00 - Ada Lovelace Centre Inria de l'université de Bordeaux 200 Av. de la Vieille Tour, 33405 Talence
in front of the jury composed of
- Martin HACHET - Directeur de recherche - Centre Inria de l'Université de Bordeaux - Directeur de these
- Thierry DUVAL - Professeur des universités - IMT Atlantique - Rapporteur
- Anne-Hélène OLIVIER - Maîtresse de conférences - Université Rennes 2 - Rapporteur
- Hélène SAUZéON - Professeure des universités - Université de Bordeaux (détachée à Inria) - Examinateur
- Anastasia BEZERIANOS - Professeure des universités - Université Paris-Saclay - Examinateur
- Arnaud PROUZEAU - Chargé de recherche - Inria Saclay - CoDirecteur de these
Today, the production of knowledge and data is increasing in many fields in order to analyze, comprehend, and explain phenomena that are sometimes complex. Communicating this knowledge to non-expert audiences can sometimes be complicated, and conventional modes of transmission based on text, graphics or audio-visual reports can reach their limits. Immersive technologies such as virtual reality offer new opportunities in visualization by immersing users in interactive environments. Virtual reality can benefit from this interactive immersion to involve users in the process of understanding phenomena more easily than conventional modes of knowledge transmission could. However, introducing relevant educational virtual reality experiences to a novice audience involves several challenges. My thesis therefore focuses on addressing these challenges. Firstly, we investigate how novice users learn how to use virtual reality applications. Secondly, we explore a new type of empathetic data visualization in which the user is immersed in the heart of the data. Finally, we study the immersive authoring of interactive experiences to favor the democratization of the creation of interactive content by non-expert users. Overall, this thesis aims to contribute to the emergence of new types of visualization through the study and design of immersive interactive visualization experiences for non-experts.
ED Sociétés, Politique, Santé Publique
Joint models with heteroscedastic residual variance: application to the study of the impact of blood pressure variability on competitive health events.
by Léonie COURCOUL (Bordeaux Population Health Research Center)
The defense will take place at 14h00 - Amphi Louis BPH - Université de Bordeaux - Campus Carreire, 146 Rue Léo Saignat, 33000 Bordeaux
in front of the jury composed of
- Hélène JACQMIN-GADDA - Directrice de recherche - Université de Bordeaux - Directeur de these
- Catherine LEGRAND - Professeure - Louvain Institute for Data Analysis and Modeling Institute of Statistics, Biostatistics and Actuarial Sciences - Rapporteur
- Jérémie GUEDJ - Directeur de recherche - UMR 1137, Inserm and Université Paris Cité - Rapporteur
- Nicola COLEY - Epidémiologiste - CHU de Toulouse - CERPOP (INSERM-Université Toulouse III) - Examinateur
- Cécile PROUST-LIMA - Directrice de recherche - Université de Bordeaux - Examinateur
This work aims to develop joint models that rigorously account for individual variability in a longitudinal marker as a risk factor for health events. These joint models combine a location-scale linear mixed model to describe the evolution over time of a biomarker, and a survival model for one or two (semi-)competitive events in which the event risk(s) is(are) adjusted for the current value, slope and variability of the biomarker. The location-scale linear mixed model is a mixed model in which the residual variability is defined based on individual random effects and/or covariates. In this work, this results in individual residual variability that may depend on time and covariates, or allows for differentiating between inter-visit and intra-visit variances when multiple measurements are collected at each visit. In the first part, we propose a joint model for competing risks with time-dependent residual variance. Applied to the PROGRESS clinical trial, this model allows the study of the impact of current blood pressure variability on the risks of cardio and cerebrovascular events, while accounting for the competing risk of death. In the second part, we present an illness-death joint model that distinguishes between inter-visit and intra-visit residual variabilities and for an interval-censored time-to-event, competing with a terminal event, such as death. This model is applied to the Three Cities cohort to evaluate the impact of inter-visit and intra-visit blood pressure variability on the risk of dementia and death. The third part describes an R package developed to allow users to estimate various joint models from this work. These models differ depending on how residual variability is modeled and the type of survival model used. The latter can handle on or two (semi-)competitive events, interval censoring or delayed entry. This package also offers tools for graphical evaluation of model fit to the data and dynamic prediction of health event risks based on model estimates.