Doctor: Rabia Azzi
Title: Contribution of the semantic web to improve the prevention of cardiovascular diseases
Supervisor: Sylvie Despres
Doctoral school: ED 146 Doctoral school Galilée, University Sorbonne Paris Nord
Date of thesis defense: 02/2020
Jury: Jérôme Nobecourt, Pierre Meneton, Gayo Diallo.
Cardiovascular disease (CVD) and cancer are the leading cause of death and morbidity in men and women in France, and their annual cost is high. The prevention of these chronic diseases is, with their early detection and their rapid and effective management, a possible way to reduce this cost. This prevention involves identifying the modifiable risk factors associated with these chronic diseases, among which are diet and physical inactivity. Thus, the National Health Nutrition Program was set up in France, helping the French to have a better diet, to help reduce the incidence of these diseases. The objective of this thesis is the construction of a system of personalized suggestions based on the profile of the individual and his cardiovascular risk. This approach requires the establishment of an interdisciplinary approach involving researchers in the fields of computer science, epidemiology and nutrition. The importance of this collaboration is justified by the need to produce suggestions supported by proven research in these areas. The first contribution of this thesis is the integration of semantic web technologies into a new approach to cardiovascular risk assessment that takes into account the interactions between these factors. The proposed approach consists of : (i) extracting and exploiting knowledge from statistical presentations of results ; (ii) construct a conceptual model based on this knowledge; (iii) dynamically visualize the resulting model to better understand the cascading effects. The creation of the visualization tool MCVGraphViz allowed to implement this strategy. The second contribution consists in proposing a solution to exploit the knowledge present in the health plans and the recommendations concerning the prevention of cardiovascular diseases in France. Thus, we opted for a modular approach integrated into the tool MCVGraphViz that allows to produce recommendations (diet, physical activity, etc.) based on the assessed cardiovascular risk and the profile of the individual (sensory preferences, allergic constraints, physical capacity, etc.). The third contribution concerns the nutritional qualification of cooking recipes for a better follow-up : the approach is based on techniques of automatic language processing and ontological reasoning to qualify a nutritional point of view of cooking recipes. Many perspectives are exposed. Most of them aim to improve referral systems and the expressivity of the cardiovascular disease knowledge base.