RSNA 2012 

Abstract Archives of the RSNA, 2012


LL-INS-SU7B

RadLex® Indexing of a Large Collection of Biomedical Images

Scientific Informal (Poster) Presentations

Presented on November 25, 2012
Presented as part of LL-INS-SU: Informatics Lunch Hour CME Posters  

Participants

Charles E. Kahn MD, MS, Presenter: Shareholder, Hotlight Inc Officer, Hotlight Inc

CONCLUSION

The RadLex® vocabulary provides a uniform terminology for clinical practice, research, and education in radiology. Almost all of the images in a large collection images from peer-reviewed biomedical journals could be annotated with at least one RadLex® term, and more than half of the images were annotated with six or more terms. The automated annotation process has enabled the development of a service to illustrate RadLex® terms with related images.

BACKGROUND

RadLex® is an ontology (knowledge model) of standardized terms for clinical practice, research, and education in radiology. We sought to apply RadLex® to annotate a large database of biomedical images.

DISCUSSION

Of the 251,613 images in the ARRS GoldMiner® collection as of March 2012, 235,371 (93.5%) were annotated with at least one RadLex term, 138,416 images (55.0%) were annotated with six or more RadLex® terms, and 5,996 images (2.4%) were annotated with 20 or more RadLex® terms. Of the 34,895 terms in the current version of RadLex® (version 3.5, released 7 Oct 2011), 6,954 (19.9%) were mapped to images in the collection. A web service to identify images has been integrated into the RadLex® web site (www.radlex.org) to display related images for each term.

EVALUATION

The National Center for Biomedical Ontology (NCBO) Annotator web service was used to discover RadLex® terms in the captions and article titles of all figures indexed by the ARRS GoldMiner® radiology image search engine, which indexes images from 262 peer-reviewed biomedical journals. We analyzed the number of annotations per image and the number of appearances per RadLex® term. We constructed a web service to return images for a specified RadLex® identifier and a standalone web page to retrieve images by RadLex® term.

Cite This Abstract

Kahn, C, RadLex® Indexing of a Large Collection of Biomedical Images.  Radiological Society of North America 2012 Scientific Assembly and Annual Meeting, November 25 - November 30, 2012 ,Chicago IL. http://archive.rsna.org/2012/12022801.html