You are here
Identifying Anatomical Concepts in Biomedical Text for Automatic Selection of Images
We discuss a method for automatically suggesting appropriate illustrations for anatomically-oriented text. The system uses natural language processing and domain knowledge in the Unified Medical Language System (UMLS) and relies on 'part of' information in an anatomical ontology. We discuss the application of our methodology to a set of gastrointestinal endoscopy (GIE) reports. Rather than providing a separate image for each anatomical concept in a report, our system determines the parts of coherent structures being discussed.