RadLex Effort Targets Imaging Data Variability

Inconsistent naming of imaging series can hinder radiology workflows, from unreliable hanging protocols to difficulty automating image selection and analysis. A new initiative through RadLex, RSNA’s controlled radiology lexicon, aims to address these challenges by developing standardized imaging series naming conventions to reduce longstanding variability across imaging systems, vendors and institutions.

Series descriptions are often stored as free text, resulting in wide variation across vendors, institutions and even identical protocols performed on different scanners. “There are a million different ways that the same information could be represented, making it difficult to programmatically identify a particular series,” said John Mongan, MD, PhD, associate chair for translational informatics and professor of clinical radiology in the Department of Radiology and Biomedical Imaging at the University of California, San Francisco.

Widely used in standardized radiology exam naming, RadLex is now expanding to address variability at the imaging series level. Differences in series naming can directly affect clinical workflows.

“Hanging protocols frequently fail to display images consistently, no matter what vendor PACS you have, because naming conventions differ across systems,” said Audrey Verde, MD, PhD, neuroradiologist, chair of the RSNA RadLex Committee, and chair of Logical Observation Identifiers Names and Codes (LOINC) Radiology.

“Integrating new sites often requires extensive remapping of naming conventions,” added Stacy O’Connor, MD, MPH, MMS, a professor of radiology in abdominal imaging and vice chair of innovation in the Department of Imaging Sciences at the University of Rochester Medical Center in New York.

“The life cycle of a radiology exam depends on seamless data handoffs across multiple systems,” Dr. Verde said. “When data varies across sites and modalities, those breakdowns can affect the radiologist workflow and ultimately patient care.”

As imaging volumes grow and systems expand, the impact of inconsistent naming becomes more pronounced. “As imaging entities grow in complexity, the accumulation of variability across scanners, vendors and institutions becomes a tsunami,” said Greg Zaharchuk, MD, PhD, a professor of radiology in neuroimaging and neurointervention at Stanford University in California. “It undermines otherwise well-designed systems and limits adoption of tools that could improve efficiency and accuracy.”

Toward More Consistent Imaging Data

While there is broad agreement on the need for standardization, reaching consensus on how to define and structure series naming is more complex. “There may be hundreds or thousands of equally valid approaches, reflecting differences in personal and institutional preferences,” Dr. Mongan said. “These variations make it difficult to establish a single approach that works across settings.”

Standardizing series naming could enable more consistent downstream use of imaging data across clinical workflows, research and AI applications. “With standardized naming, downstream processes can be automated, improving efficiency and reducing manual effort.” Dr. O’Connor said.

Standardized naming also allows systems to reliably identify relevant prior studies, apply appropriate report templates and support consistent hanging protocols. “At an organizational level, the benefits extend beyond workflow improvements,” Dr. Verde said. “While standardization requires upfront investment, the opportunity for efficiency is substantial, supporting better use of data for clinical care, research and operation.”