| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Editorial |
1 From the Department of Radiology, University of Pennsylvania Health System, Radiology Administration, Penn Tower Lobby Level, 399 S 34th St, Suite 100, Philadelphia, PA 19104. Received September 12, 2006; revision requested and received September 14; accepted September 14. The author is shareholder in eDictation and a consultant for Elsevier. Address correspondence to the author (e-mail: langlotc{at}uphs.upenn.edu).
For several decades, the American College of Radiology’s Index for Radiological Diagnoses (known as the ACR Index) (1) has served admirably as an indexing system for radiology teaching files. The ACR Index was originally developed to categorize and organize the image-based interesting cases collected by radiologists, often in paper folders on office shelves. As radiology clinical practice and radiology education move online, there is an increasing need for an indexing system that works equally well in the digital world. The Radiological Society of North America’s (RSNA’s) project RadLex is designed to address that need. This editorial discusses the ACR Index and introduces the RadLex lexicon, a new method for indexing online educational materials for radiologists and educators.
The ACR Index
The ACR Index has several attractive features for indexing image-based teaching files. First, it offers both anatomic and pathologic identifiers. By convention, ACR Index codes are decimal numbers. The numeric code for the anatomic location appears before the decimal point, and the numeric code for the pathologic entity appears after the decimal point. For example, the code for primary adenocarcinoma of the lingula is represented by the ACR code 642.3212.
Second, the ACR Index was created to be used by humans. Each digit of the ACR Index code denotes a more specific term in the taxonomy. For example, lung, mediastinum, and pleura are represented by 6; the left upper lobe is represented by 64; and the lingula, a subdivision of the left upper lobe, is represented by 642. Likewise, neoplasms and neoplastic-like conditions are represented by 3; primary malignant neoplasms are represented by 32; carcinoma-type neoplasms are represented by 321; and primary adenocarcinoma is represented by 3212. This numbering scheme enabled many radiologists to remember the codes for familiar anatomic locations and diagnoses that they commonly catalogued in their personal image libraries (Table).
|
Shortcomings of the ACR Index
As teaching materials and other personal image libraries have moved online, some of the attributes of the ACR Index that made it desirable for organizing image-based teaching materials have become substantial limitations (2,3). Computers now have the capacity to manage huge lists of index terms and their interrelationships. These online lexicons are valuable tools to index online content. For example, SNOMED-CT (Systematized Nomenclature of Medicine–Clinical Terms, from the College of American Pathologists, Northfield, IL) contains over 350,000 concepts that can be used to organize the information in electronic medical records. Because the ACR Index contains only a few thousand unique terms, it is easier for humans to remember, but it offers much less detail than other terminology systems, including SNOMED-CT and the Foundational Model of Anatomy (4).
A second drawback of the ACR Index is its numbering scheme, which was designed for the information processing limitations of humans. The ACR Index numbering scheme uses digits to represent logical relationships between terms. For example, since 64 signifies the left upper lobe, it is immediately apparent that 642 must represent one of the parts of the left upper lobe. This fixed relationship between digits and concepts sometimes makes it difficult to add or retire terms without changing the codes of other nearby concepts. Human limitations also require that the numeric codes be brief. The ACR Index avoids long numeric codes that are difficult to remember by restricting detail. For example, the ACR Index does not provide a code for the apical segment of the right lower lobe, or for most other lung segments. To address this issue, most modern terminology systems completely hide the numeric identifiers from human users, since software systems can be designed to interact with the unique identifiers in the background.
Indexing and Retrieval of Electronic Images
To help usher in the era of electronic teaching materials, the RSNA has developed the Medical Imaging Resource Center (MIRC) (5,6), a set of tools that enables users to connect electronic teaching files to one another over the Internet. The MIRC project offers an online authoring tool for creation of electronic teaching files and other forms of annotated personal image libraries. As the shift to online educational materials has occurred, there is a clear need for a more complete and computer-friendly index to describe imaging findings. Unfortunately, no extant medical terminology systems can meet the needs of online radiology indexing, typically because these systems do not contain a complete set of imaging terms (7,8).
The RadLex Project
To address this limitation, the RSNA has embarked on the RadLex project, which is designed to fill in the gaps in other medical terminology systems, thereby creating a single source for medical imaging terminology. One of the primary goals of this project is to create a terminology that can be used to annotate, index, and retrieve content from MIRC. The RSNA was fortunate to have the support of the ACR, which contributed the ACR Index to the RadLex project, thereby ensuring that image-based teaching materials coded with the ACR Index can be easily indexed with RadLex codes. To avoid duplicate effort, RSNA also has arranged with the College of American Pathologists to use a subset of SNOMED-CT terms as a starting point for the RadLex lexicon (9). Finally, the RSNA has enlisted the cooperation of numerous standards organizations, including Digital Imaging and Communications in Medicine (DICOM) (10) and Integrating the Healthcare Enterprise (IHE) (11), as well as a wide range of professional organizations drawn mostly from radiology subspecialties. These cooperating organizations have assisted in identifying the chairs of the RadLex subcommittees and will contribute to the dissemination of the lexicon after it is released.
Over the past year, over 90 radiologists have met in groups, including thoracic, abdominal, musculoskeletal, neurologic, cardiovascular, and pediatric imaging subcommittees, to deliberate over draft term sets. Meetings in fall 2005 focused on anatomic terms, while meetings in winter and spring 2006 dealt with imaging findings and pathologic terms. These anatomic and pathologic terms are available for public comment on the RadLex Web site (http://www.rsna.org/radlex/). Over the coming year, the RSNA will be working with numerous modality-based subspecialty organizations to create a consistent language to describe the devices, procedures, and imaging sequences used to create imaging examinations.
Comparing RadLex to the ACR Index
RadLex will differ from the ACR Index in several important ways. RadLex already contains over 8,000 anatomic and pathologic terms, many of which are not currently available in the ACR Index or in any other medical terminology system. Thus, the comprehensiveness of RadLex enables indexing of images with more specific concepts. For example, RadLex has augmented the anatomic and pathologic codes in the ACR Index with additional types of terms, including (a) the devices, procedures, and imaging techniques used to acquire radiology images; (b) the perceptual and analytical difficulty of the interpretation; and (c) the diagnostic quality of the images. Another key distinguishing feature of RadLex is that it is designed to be continuously supplemented and updated with incorporation of new concepts, including harmonization with other popular medical vocabularies and term sets, such as SNOMED-CT, ICD9 (International Classification of Diseases, 9th Revision, from the World Health Organization), CPT (Current Procedural Terminology, from the American Medical Association, Chicago, IL), BrainInfo (12), and, of course, the ACR Index. In particular, radiologists whose teaching materials are organized according to the ACR Index can rest assured that these materials can be readily and automatically updated to the RadLex system.
After the draft RadLex term sets are revised based on public comments, the first full release of RadLex will be available on the RSNA Web site at the time of the RSNA annual meeting in November 2006. RadLex will be available in a variety of forms, including a tabular format, which can be loaded directly into Excel, a variant of extensible markup language (XML) format called ontology Web language (OWL) (13), and as a database file, which can be loaded into an open-source vocabulary management tool called Protégé (14). RadLex terms also will be available for browsing or searching via the Internet or while using the MIRC authoring tool. For a sample of how a term browser might work, visit http://mirc.rsna.org/radlex/service to explore a pilot version of the RadLex thoracic lexicon.
The developers of RadLex recognize that it must continue to evolve to retain its utility. A significant ongoing effort will be required, both to ensure that new concepts are incorporated into RadLex and to maintain the cross-links between RadLex and other vocabularies when they are updated. As the first version of the new RadLex resource nears completion, we welcome your e-mail feedback at rl-feed{at}rsna.org. Please consider taking advantage of the comprehensiveness and computational versatility of RadLex when it is released this fall.
References
This article has been cited by other articles:
|
|
 |
|
  A. W. C. Kamauu, J. J. Whipple, S. L DuVall, K. M. Siddiqui, E. L. Siegel, and D. Avrin Informatics in Radiology: IHE Teaching File and Clinical Trial Export Integration Profile: Functional Examples RadioGraphics, July 1, 2008; 28(4): 933 - 945. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. L. Rubin, N. H. Shah, and N. F. Noy Biomedical ontologies: a functional perspective Brief Bioinform, January 1, 2008; 9(1): 75 - 90. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. E. Kahn Jr. and C. Thao GoldMiner: A Radiology Image Search Engine Am. J. Roentgenol., June 1, 2007; 188(6): 1475 - 1478. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOGRAPHICS | RADIOLOGY | RSNA JOURNALS ONLINE |
