An architecture for semantic integration of data and medical images

In organisations providing health services, there exist different data sources (e.g. clinical history, demographics data, DICOM files) of diverse nature, which are scattered storage and come from heterogeneous sources. Additionally, the DICOM format stores patient information and medical images. These files are managed in PACS, however PACs does not provide diagnostic support tools. In this paper, a data integration architecture oriented to enrich medical images using metadata extracted from a controlled vocabulary is presented. The architecture was instantiated in a prototype that provides image annotation mechanisms - manual and automatic – and strategies for searching and retrieving data and images using keywords or descriptors MPEG-7, which are different from traditional ones. The annotation is based on a controlled vocabulary that is part of a taxonomy of concepts and medical terms.