Information Extraction from Echocardiography Records

Electronic health records are a rich source for medical information. However, large parts of clinical diagnosis reports are in textual form and are therefore not per se usable for statistical evaluations. To transform the information from an unstructured into a structured form is the goal of medical language processing. In this paper we want to propose an approach for the creation of a training corpus for information extraction from echocardiography reports and the creation of a sequence labeler based on keyword matching and window-based disambiguation. The outcomes presented in this paper are the first results from ongoing work from a series of medical projects.