Medical reports are, quite often, written and stored in computer systems in a non-structured free text form. As a consequence, the information contained in these reports is not easily available and it is not possible to take it into account by medical decision support systems. We propose a methodology to automatically process and analyze medical reports, identifying concepts and their instances, and populating a new ontology. This methodology is based in natural language processing techniques using linguistic and statistical information. The proposed system was applied successfully to a set of medical reports from the Veterinary Hospital of the University of Évora.