While there's an increasing need to generate consumable data from free-text medical imaging reports, there's a lack of clinical research in this area, said presenter Dr. Robyn Cairns.
"The clinical utility of these techniques is poorly understood, and literature that validates automated encoding of clinical free-text is limited," she said. "Currently, there is a lack of standardized research techniques for clinical free-text codification and limited clinical validation of the [NCBO] Annotator Web service."
In testing, the researchers found that the NCBO Annotator -- a flexible and easy-to-use open-source tool -- can identify existing exact-match RadLex concepts in clinical free-text reports.
"However, the performance of the NCBO Annotator for correctly extracting all terms that were considered clinically relevant in pediatric orthopedic medical imaging free-text reports and excluding irrelevant terms is limited," she said.