Yi Hong MS, Presenter: Nothing to Disclose

Jin Zhang PhD, Abstract Co-Author: Nothing to Disclose

Xi Meng, Abstract Co-Author: Nothing to Disclose

Charles E. Kahn MD, MS, Abstract Co-Author: Shareholder, Hotlight Inc Officer, Hotlight Inc

This report describes a MDS visualization study of the semantic relationships among co-occurring terms in XML-encoded radiology reporting templates. It presents a new approach to represent radiological information visually.

Through the radiology reporting initiative, RSNA has made the XML-encoded reporting templates available on the RSNA Reporting Template Library website. The elements of the reporting templates have been mapped to RadLex using a semi-automated mapping tool. The multidimensional scaling (MDS) visualization technique incorporated with SPSS was used to conduct a "gap analysis" to identify unmapped reporting elements.

To conduct the MDS analysis, an unmapped element "poor venous opacification" was identified as a seed term to select related terms that co-occur with it in the designated template context. Since the SPSS is better at dealing with a smaller data set, 88 of 135 related terms were selected from the "Finding" sections of three CT exam reporting templates (CT Cardiac, CT Cardiac Bypass Graft, CT Pulmonary Veins) that include the seed term “poor venous opacification". To make it possible to be added to RadLex, the seed term was chosen as it does not exist in RadLex, SNOMED, or UMLS, and it is a term that is frequently used by radiologists.

As the elements in a XML-encoded reporting template are organized hierarchically, we may determine the relationships between the seed term and its co-occurring terms by examining how far away the terms are located in the hierarchy. The greater the distance between the two terms, the weaker is the strength of their association. MDS was chosen for the visualization study of the CT exam reporting templates because it can handle various types of data ranging, and it is good at small to medium-sized data sets like the reporting templates data set. The final stress value is 0.01268 and the RSQ value is 0.99957, which shows a good result of the MDS analysis on association of reporting terms.

Hong, Y, Zhang, J, Meng, X, Kahn, C, Using a Visualization Technique to Represent Information in Radiology Reporting Templates.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11010037.html