Semantic-aware epidemiological surveillance system : application to tuberculosis in cameroon

Abstract

Epidemiological surveillance systems implement complex processes for the collection, analysis and interpretation of health data, for the planning or the evaluation of public health practices and policies. These systems must be able to provide up-to-date, precise and complete information to stakeholders whose interests are diverse and evolve with time. There are many difficulties faced when putting in place such systems. For example, the ability for such systems to collect, transmit and manage structured data, while taking into consideration the security, authentication and confidentiality of the data is crucial. To this end, a software editor known as "IMOGENE", based on Model Driven Architecture (MDA) was developed by MEDES in Toulouse. A joint project led by UMMISCO, MEDES, Centre Pasteur du Cameroun and National Program to Fight against Tuberculosis, made use of the latter platform to develop and deploy a tuberculosis surveillance system named EPICAM. This project showed that the absence of semantic links between data only allowed the exploitation of information explicitly defined in a database. The idea presented in this thesis to solve this problem is the use of information contained in source code, to infer new knowledge and integrate them in a domain ontology. To be precise, we propose a solution based on Hidden Markov Models (HMMs), which as opposed to other existing techniques that are limited to extraction of terminologies, concepts and properties also enables learning of axioms and rules. The implementation on the source code of the EPICAM platform has allowed us to describe in a clear, precise and succinct manner what we consider as principal information obtained, which has been evaluated and validated by domain experts.