Beverly Collins, PhD

Over the last several decades, radiologists have made incredible advances in noninvasive imaging examinations, and in transmitting, storing, and processing the increasingly complex images produced by these exams. Despite the stunning progress in producing and managing images, there have been no analogous dramatic changes in our ability to communicate clearly and consistently the results of an imaging procedure.1

The primary final product of most radiology examinations still is a text report that employs nonstandardized terminology, that is loosely organized, and that is difficult to compare to reports from other radiologists at other institutions.2 For example, radiologists trained at one institution may use the term “infiltrate” to describe a nonspecific pulmonary parenchymal opacity on a chest radiograph, whereas those trained at another site may use the same term to denote a pneumonia. Likewise, some radiologists may use the term “bronchiolitis obliterans organizing pneumonia (BOOP),” whereas others may use “cryptogenic organizing pneumonia” to describe the same phenomenon. Examination names often are similarly confusing. For example, the difference between a brain MRI and a head MRI (if any) may vary among radiology practices. And we are all familiar with the variety of proprietary terms to describe the same common MRI pulse sequences. The lack of a consistent language limits our ability to search, retrieve, compile, and analyze images, whether they are in picture archiving and communications system (PACS) archives, teaching files, or research databases.

THE RADLEX PROJECT

To respond to the need for consistency, the Radiological Society of North America is undertaking the development of a unified radiology lexicon, called RadLex.3 The project draws support from the ACR and several radiologic subspecialty professional societies, whose members have volunteered to serve on lexicon development committees. The goal of the RadLex project is to create a freely available uniform terminology for capturing, indexing, and retrieving information in a variety of radiology information sources. The result will be a single source for terminology that can be used to describe salient aspects of an imaging examination including modality, technique, visual features, anatomy, findings, and pathology. Rather than duplicate the efforts of others, RadLex will unify and supplement radiology terms in other medical lexicons. The new terminology will be made available to all users on the Internet with cross references to these other widely used lexicons and standards.

WHO WILL BENEFIT FROM RADLEX?

This new lexicon will transform the indexing, query, and retrieval of images for teaching, clinical care, and research. Examples of each benefit are discussed below.

Radiology Educators and Students. In combination with teaching file software developed by RSNA’s Medical Image Resource Center (MIRC) project,4 RadLex will enable rich indexing of online teaching materials. For instance, consider a teaching case illustrating the typical appearance on contrast-enhanced CT of a primary adenocarcinoma arising from the posterior basal segment (S10) of the lower lobe of the left lung. The tumor invades the left inferior pulmonary vein. Currently, this case would be indexed using ACR code 65.3212 (left lower lobe/malignant neoplasm-primary adenocarcinoma). Online queries of such teaching files could select cases using only the anatomy and pathology terms in the ACR Index.

RadLex will incorporate the ACR Index and maintain compatibility with it. But RadLex also will provide many more categories of terms in addition to anatomy and pathology. For example, the following RadLex terms are available for indexing and search of this teaching file: posterior basal segment (S10) of the lower lobe of the left lung, adenocarcinoma, left inferior pulmonary vein, computed tomography, contrast-enhanced, basic perceptual difficulty, intermediate analytical difficulty. Because RadLex will be linked to the ACR Index, the RadLex anatomy and pathology equivalents of 65.3212 can be found quickly. Similarly, RadLex will be linked to the Unified Medical Language System, so other analogous terms can be found readily in other terminology and coding systems, such as ICD-9 and CPT4. Once files are indexed with RadLex, teaching materials are more likely to be found and used by students, trainees, and radiologists seeking continuing education opportunities.

Clinical Radiologists. Consider a radiologist interpreting a chest CT showing a tree-in-bud appearance. The radiologist is unsure whether the examination being interpreted truly exhibits this feature, and does not know the diagnostic possibilities that might explain the appearance. Rather than spending time sifting through textbooks, course syllabi, and recent journal articles, and perhaps also consulting PubMed5 and Google online, a radiologist will be able to search for the tree-in-bud term on the RadLex web site. A page detailing that term will link to representative images that match the case at hand. Because RadLex will provide a standard for the communication of radiology knowledge, PACS vendors could provide a “gamuts on demand” button triggered by RadLex terms. The radiologist selects that button on the workstation and enters “tree-in-bud.” A differential diagnosis of tree-in-bud appearance is displayed, including links to relevant full-text articles from journal web sites.

Radiology Researchers. RadLex can help to accelerate the rate at which new technologies are evaluated. Consider research teams at three different research institutions, each of which has an interest in studying the ability of shoulder MRI to predict the outcomes of shoulder surgery. Each team is designing a database to collect key information for exploratory research. Currently, each institution designs forms that collect different information, and may ask the questions for similar information in different ways. For example, each form uses a different vendor’s proprietary terms to describe the MRI pulse sequences. When each research team has completed its study, the data cannot be pooled or compared readily.

Curtis P. Langlotz, MD, PhD

RadLex will enable each research team to find common preferred terms to describe anatomy, findings, and MRI related to shoulder pathology and MRI pulse sequences. These terms enable the information in each research database to be collected in a consistent form. As a result, experimental data and results are easier to compare and combine once the studies are concluded. The three institutions can pool their data to provide stronger conclusions than would otherwise be drawn from three individual disparate studies. This type of shared data system is one of the key goals of the Roadmap Initiatives in Bioinformatics and Computational Biology at the National Institutes of Health and of the Common Data Elements initiative at the National Cancer Institute.

EXISTING SHORTCOMINGS

Before the RadLex group decided to create a new lexicon, it carefully considered whether existing terminology systems, which are widely used in electronic medical record systems, contain the terms that radiologists commonly use. (See Table 1 for a list of popular terminology systems.) Researchers have found that only 40% to 60% of terms found in radiology reports are contained in the most comprehensive medical vocabularies.1 Another study found that two of the most comprehensive sources of terminology, the UMLS Metathesaurus and SNOMED International (an earlier version of SNOMED-CT), contain only 14% and 23%, respectively, of the terms needed to capture the meaning of an ultrasound imaging report.

Classification and terminology systems developed specifically for the field of imaging are also not adequate to the needs of today’s radiologists, with a few notable exceptions. The ACR Index was helpful for the organization of an individual radiologist’s physical files, but limited (covering only 19% of ultrasound concepts), and is not optimized for online indexing and searching. Some imaging subspecialties have developed standard terms, including a chest imaging glossary,6 a glossary of MRI terms,7 a lumbar disc nomenclature,8 a tumor ablation terminology,9 and a CT colonography reporting and data system.10 While these efforts have been extremely helpful in standardizing the way radiologists communicate about certain clinical problems, these groups of terms cover only a small portion of radiology practice, and are not readily amenable to computer processing.

A notable exception is the ACR Breast Imaging Reporting and Data System (BI-RADSTM),11 which has transformed mammogram interpretation. Spurred in part by legislation requiring the assignment of codes to mammograms and the compilation of performance audits for radiologists, BI-RADS is now an integral part of breast imaging practices. It is used as a succinct method of teaching how to interpret mammograms; it has enabled direct comparison of breast imaging research results; and it has resulted in numerous programs to measure and improve the performance of radiologists. These are some of the benefits that RadLex may spread to other radiology subspecialties.

Terminology systems

Source

SNOMED-CT (Systematized Nomenclature of Medicine-Clinical Terms)

College of American Pathologists

UMLS (Unified Medical Language System)

National Library of Medicine

NCI Thesaurus

National Cancer Institute

LOINC (Logical Ovservation Identifiers, Names, and Codes)

Regenstrief Foundation

CPT4 (Physicians’ Current Procedural Terminology)

American Medical Association

ICD-9-CM (International Classification of Diseases: 9th revision, Clinical Modification)

Health Care Financing Administration


WHAT TO EXPECT FROM RADLEX

The RadLex effort has been led and coordinated by a Steering Committee whose members have a special interest in radiology informatics and medical terminology. The Steering Committee began by devising a logical organization for imaging terms suitable for all radiology subspecialties. Next, a pilot project in thoracic imaging was undertaken, in collaboration with members of the Fleischner Society and the Society of Thoracic Radiology. The results of that effort can be viewed on the Internet by visiting http://mirc.rsna.org/radlex/service/ .

The RadLex effort is now expanding to address the remaining subspecialty areas, including musculoskeletal, neuroradiology, abdominal imaging, cardiovascular, pediatric radiology, and radiology examinations and protocols. This summer, the RadLex project began recruiting members for these organ system committees. Each of these efforts will be conducted in collaboration with the relevant subspecialty organizations. The RSNA is currently discussing relationships with other terminology and standards organizations, including the College of American Pathologists and DICOM. Version 1.0 of RadLex, a new and comprehensive lexicon, is expected to be available in time for the RSNA’s 2006 annual meeting. For more information on the project, see http://www.rsna.org/radlex/

Beverly Collins, PhD, is the RadLex Coordinator for the RSNA and a scientific writer and special projects manager in the Department of Radiology at the University of Pennsylvania, Philadelphia.

Curtis P. Langlotz, MD, PhD, is associate professor and chief of Health Services Research at the University of Pennsylvania, and chair of the RSNA RadLex Steering Committee.

References:

  1. Langlotz CP. Automatic structuring of radiology reports: harbinger of a second information revolution in radiology. Radiology. 2002;224:5-7.
  2. Sistrom CL, Langlotz CP. A framework for improving radiology reporting. Journal of the American College of Radiology. 2005;2:159-167.
  3. Available at: mirc.rsna.org/ radlex/service. Accessed August 4, 2005.
  4. Available at: www.rsna.org/ mirc/. Accessed August 4, 2005.
  5. Available at: www.ncbi.nlm. nih.gov/PubMed/. Accessed August 4, 2005.
  6. Austin JH, Muller NL, Friedman PJ, et al. Glossary of terms for CT of the lungs: recommendations of the Nomenclature Committee of the Fleischner Society. Radiology. 1996;200:327-31.
  7. American College of Radiology. Glossary of MRI Terms. Reston, Va: American College of Radiology.
  8. Fardon DF, Milette PC. Nomenclature and classification of lumbar disc pathology. Recommendations of the Combined Task Forces of the North American Spine Society, American Society of Spine Radiology, and American Society of Neuroradiology. Spine. 2001;26(5):E93-E113.
  9. Goldberg SN, Grassi CJ, Cardella JF, et al, and the Society of Interventional Radiology Technology Assessment Committee. Image-guided tumor ablation: standardization of terminology and reporting criteria. J Vasc Interv Radiol. 2005;16:765-778.
  10. Zalis ME, Barish MA, Choi JR, et al, and the Working Group on Virtual Colonoscopy. CT colonography reporting and data system: a consensus proposal. Radiology. 2005;236:3-9.
  11. American College of Radiology. Illustrated Breast Image Reporting and Data System (BI-RADSTM). Reston, Va: American College of Radiology.