Rationale and Objectives
RadLex was developed to create a unified language for radiologists. Despite the large number of terms, little research has evaluated the degree to which RadLex contains terms frequently used in clinical practice. The purposes of this project are to estimate the completeness of RadLex in the chest radiography domain and to characterize the absent terms. We chose chest radiography because it is a common exam generating a large number of reports, and the terms used represent a relatively well-circumscribed set of terms compared to other anatomic regions and modalities.
Materials and Methods
We collected a random sample of 100 chest radiograph reports from 1 month of routine clinical practice of three board-certified radiologists. We parsed each report’s findings and impression sections into individual objects. An “object” was defined as any discrete physical object, body part, observation, descriptive modifier, diagnosis, or procedure. Objects were compared to RadLex by entering the object into the RadLex Term Browser. We calculated descriptive statistics and compared the match rate across RadLex categories.
Results
We identified 339 unique objects, with an overall match rate of 62%. The match rate for each category was anatomic object, 77%; physiological condition, 73%; physical object, 65%; imaging observation, 47%; procedure, 0%; other, 41% ( P < .0005).
Conclusions
Our study shows that despite the large number of terms in RadLex, terms are still absent and complexities in the definitions of terms exist. However, increasing the completeness and refining the definitions in RadLex is easily surmountable, possibly using manual methods.
An ontology is a way of representing the terms and relationships in a domain. Ontologies are useful representations of domains because they are can be browsed by humans and processed by machines . RadLex , a radiology-specific ontology, was developed in 2006 to create a unified language for radiologists. A radiology-specific lexicon was needed to make more efficient use of the growing amount of electronic information in the radiology environment , in particular in the creation of electronic teaching materials, and to more accurately search reports and perform data-mining tasks. The RadLex terms were originally gathered from existent sources including the American College of Radiology (ACR) Index, SNOMED-CT, and the Foundational Model Anatomy, and now includes more than 30,000 terms . More recently, a large number of unique imaging signs were added to the RadLex ontology .
Despite the large number of terms within the RadLex ontology, little research has evaluated the degree to which RadLex contains terms frequently used in clinical practice. The single study that specifically evaluated which terms contained in clinical free text reports are in RadLex evaluated an early version of RadLex (January 2007). This study found that 84% of thoracic computed tomography (CT) terms used at a single institution and derived from an automated term identification algorithm were included in RadLex and that anatomic terms and imaging findings were the most robust sections within RadLex. Since that time, newer versions of RadLex have been released, and continuous updates to the ontology are performed. More recent research analyzed the number of terms that match reporting templates developed by the Radiological Society of North America reporting initiative and found a 41% exact match rate for terms in RadLex.
Get Radiology Tree app to read full this article<
Materials and methods
Exam Selection
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Radiograph Object Parsing
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Comparison to RadLex Term Browser
Get Radiology Tree app to read full this article<
Table 1
Analysis Categories Based on the RadLex Ontology
Category Name RadLex Ontology Categories Included Anatomic object Anatomical entity, RadLex location Physiological condition Physiological condition Physical object Object, nonanatomical substance Imaging observation Imaging observation, size or quantity, morphological characteristic, RadLex extent, RadLex distribution, RadLex temporal modifier Procedure Procedure Other Imaging procedure attribute, property, RadLex modifier/entity, other
Get Radiology Tree app to read full this article<
Statistical Analysis
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Results
Get Radiology Tree app to read full this article<
Table 2
Frequency and Match Rate of Identified Objects
Category Individual Objects Unique Objects Objects in RadLex Match Rate (%) Anatomic object 692 113 87 77 Physiological condition 288 70 51 73 Physical object 83 34 22 65 Imaging observation 465 78 37 47 Procedure 14 10 0 0 Other 155 34 14 41 Total 1697 339 211 62
Table 3
Most Frequent Unmatched Objects, in Order of Frequency
Anatomic Object Physiological Condition Physical Object Imaging Observation Procedure Other Cardiothoracic ratio Obesity Catheter tip Unchanged Anterior chest surgery Compatible with Lower Infiltrate Infusion port No other abnormality Coronary artery bypass Except for Lung base/s Anterior wedging PICC tip No other change Mammoplasty Excluding Cavoatrial junction Pulmonary vascular congestion Coronary artery stent Tiny Aortic valve replacement Lung technique Above Deep breath Hickman catheter tip Appearing since Arthroscopic surgery Correlate clinically
Get Radiology Tree app to read full this article<
Discussion
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Multiple Definitions
Get Radiology Tree app to read full this article<
Unexpected Definition
Get Radiology Tree app to read full this article<
Physiological Condition Not Represented
Get Radiology Tree app to read full this article<
Procedures
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Conclusions
Get Radiology Tree app to read full this article<
References
1. Rubin D.L.: Creating and curating a terminology for radiology: ontology modeling and analysis. J Digit Imaging 2008; 21: pp. 355-362.
2. Radiological Society of North America (RSNA). RadLex, versions 3.1-3.5. http://www.rsna.org/Informatics/radlex.cfm . Accessed January 14, 2012.
3. Langlotz C.P., Caldwell S.A.: The completeness of existing lexicons for representing radiology report information. J Digit Imaging 2002; 15: pp. 201-205.
4. Langlotz C.P.: RadLex: a new method for indexing online educational materials. Radiographics 2006; 26: pp. 1595-1597.
5. Shore M.W., Rubin D.L., Kahn C.E.: Integration of imaging signs into RadLex. J Digit Imaging 2012; 25: pp. 50-55.
6. Marwede D., Schulz T., Kahn T.: Indexing thoracic CT reports using a preliminary version of a standardized radiological lexicon (RadLex). J Digit Imaging 2008; 21: pp. 363-370.
7. Hong Y., Zhang J., Heilbrun M.E., et. al.: Analysis of RadLex coverage and term co-occurrence in radiology reporting templates. J Digit Imaging 2012; 25: pp. 56-62.
8. Wheeler P.S., Simborg D.W., Gitlin J.N.: The Johns Hopkins radiology reporting system. Radiology 1976; 119: pp. 315-319.
9. Eng J., Eisner J.M.: Informatics in Radiology (infoRAD): radiology report entry with automatic phrase completion driven by language modeling. Radiographics 2004; 24: pp. 1493-1501.