| dc.description.abstract | The huge volume of data produced in the course of research and medical practice requires the use of computerized systems. Moreover, a technical implementation of these softwares requires effective systems to "understand" the "semantic meaning" of the messages, store records, ask for specific information, meet other requests for information, besides being actually useful for a given medical informatics use case. The present study provides a thorough research on the representation of patient data through information models and biomedical ontologies, in order to identify linking points between a robust methodology, such as the ontological realism, and a flexible methodology, such as health information models. The objective of the research is to evaluate the ontological representation of clinical data, by analyzing the knowledge structure of clinical health information models. The literature review covered the state of the art of interoperability standards in medicine, ontologies as artifacts of computer systems and as a philosophical discipline, language and communication. On the empirical stage, we studied clinical models of medical data representation (archetypes) and sentential fragments of records, using the ontological realism as a framework to create a representation of ontological models and statements about patients. The methodology consisted of three steps: 1) theoretical framework creation, 2) ontological analysis of the openEHR archetype model; 3) ontological representation of medical information. The framework, based on the three worlds theory by Karl Popper, recognizes the existence of four types of representational entity: ontological entities, epistemological entities, informational records and reasoning processes. The openEHR archetype model analysis demonstrated that there are limits to the exclusive realistic representation of medical records, but also that this philosophically sound analysis significantly reduces the ambiguity in representation. Finally, the vast majority of fragments was adequately represented in the ontology. We were able to represent and retrieve information properly through medical ontologies represented in description logics. However, there are practical limitations which must be observed and treated in accordance with the technology used and usage needs. Several difficulties were encountered during the conversion of clinical data into an ontological representation. The more immediate problems relate to the treatment of different types of complex data by realist ontologies. We found limitations inherent to the technology (description logics based implementations) and even the inherent flexibility of natural language interpretation by humans, which is still far from being mimicked by automated systems. | |
