DICOM SR in Radiology: Why Automation Is Harder Than Expected
What is DICOM SR in radiology?
Imagine a radiologist reviewing a mass for a renal ultrasound and confirming its location, size, and characteristics. With DICOM SR (Structured Reporting), these numeric measurements are sent directly to the reporting system, so there’s no need to dictate them.
DICOM Structured Reporting (DICOM SR) is an imaging informatics standard that allows imaging systems to transmit measurements, observations, and calculations as structured, machine-readable data, rather than free text, alongside DICOM images. In radiology, it’s used for exams such as ultrasound and DEXA, which often involve many measurements.
However, most imaging systems today implement only a limited subset of DICOM SR, typically transmitting numeric measurements and a small number of parameters, necessitating additional dedicated reporting packages to capture observations and perform calculations. Because of this limit, sonographer worksheets are scanned and added to a study. The need for a holistic, streamlined process is evident in the daily inefficiencies clinicians experience.
How does DICOM SR improve radiology reporting efficiency?
DICOM SR makes reporting more efficient by eliminating the need to dictate measurements, reducing transcription errors, and automatically populating report values. Departments typically see measurable improvements in KPIs, including shorter report turnaround times, shorter dictation durations, and fewer transcription errors.
When properly implemented, it reduces the time sonographers spend on image acquisition documentation while also improving note quality and consistency. Additionally, it lets radiologists spend more time interpreting key images and less time dictating.
Radiologists with extensive use of DICOM SR report more than a minute of dictation time per exam. Dr. Steven Horii reported a 40% time savings in exams with many measurements [Link]. Furthermore, speech recognition system errors affect 4.23% of studies, with some being ‘significant’ or ‘very significant’ [Link].
Why is DICOM SR more complex to implement in radiology than in other specialties?
Radiology lacks standardized DICOM SR reporting templates.
Unlike cardiology, vascular, or obstetrics—where well-defined SR templates exist—radiology measurements vary across vendors, exam types, and software versions. Because of this, simple one-size-fits-all SR mapping does not work in radiology.
Why do DICOM SR mapping requirements differ between ultrasound vendors?
Since there are no well-defined DICOM SR templates for radiology exams, each ultrasound company uses its own code values, code meanings, and data structures in DICOM SR. Even common measurements, such as thyroid lobe sizes, vary across vendors, so each system needs its own mapping.
One example illustrates how the lack of radiology templates has led to different content of an SR document. The figure below shows a small portion of the ‘coded concepts’ structure that is to ensure that each piece of clinical information has a unique identifier and is machine-readable. For ease of illustration, many content items, such as units, status selection (e.g., mean, latest), and value types (text strings, numeric), are omitted.
| Vendor | Concept Code | Code Meaning | Value |
| Canon | G-C0E3 | Finding Site | Anatomic Structures |
| G-C171 | Laterality | Right | |
| T-B6000 | Thyroid | ||
| G-A220 | Width | 0.73 cm | |
| GE | GEU-1005-7 | Anatomy Label | Thyroid |
| 121206 | Distance | 1.72 cm | |
| GEU-1005-5 | Measurement Label | RT LOBE H | |
| Philips | T9900-02/04 | Measurement/Label | LOBE H |
| G-C171 | Laterality | Left | |
| 121206 | Distance | 12.4 mm | |
Can DICOM SR mappings change with software updates?
Yes. Vendors can modify DICOM SR structures, codes, or storage locations when they update their software. If these changes are not updated, mappings that once worked stop transferring. A successful project needs to allocate a mapping resource to update mappings when they break.
Why do most DICOM SR projects fail to handle lesions and masses?
Radiology reporting software (PowerScribe/Fluency/other) reporting templates match each measurement to a single field (merge field or token) to achieve a structured document. But lesions and masses vary in number and location across patients, so automatic coding into templates does not work well. To handle multiple lesions, the DICOM SR parsing and reporting software needs logic to display only the measurements taken, so radiologists are not required to delete extraneous fields or select from long picklists.
An ideal solution is shown in the preliminary report below, where a right upper pole nodule is correctly placed. This minimizes radiologists’ cognitive load and maximizes interpretive time.
Why isn’t simple data parsing enough for DICOM SR automation?
Parsing can extract numerical measurements, but it does not contain other specific information, such as location or clinical observations. An Imorgon study revealed that sonographer measurements account for less than 50% of the information in a radiology findings section [Link]. Additional reporting software with logic engines have electronic worksheets with the additional tasks needed to create a preliminary report. Without this, radiologists are effectively dictating diagnostic reports manually.
What additional tasks should DICOM SR automation eliminate?
Successful automation projects should also take care of common and specific uses:
- Prior measurement retrieval
- Percent change calculations
- Volume calculations
- Clinical risk calculators (e.g., TI-RADS, O-RADS)
If the tasks above remain manual, radiologists will retain most of the cognitive burden, even if the measurements are transferred.
An example of an electronic worksheet that includes typical report content for a successful automation project
Why do IT-led DICOM SR projects often stall?
IT-led projects manage a project’s DICOM infrastructure needs. However, DICOM SR automation projects also require:
- Sonographer: coordinate with manufacturers’ application support, perform test scans, and later provide peer training.
- Radiologist: approve necessary measurements, worksheet contents, and the final reporting templates.
- Implementation Team: If an external resource is not implementing, knowing how to ask all of the small clinical decisions, such as which measurements to include, how many lesions to support, and how to integrate conditional data into radiology reporting templates.
- IT Team: Even accounting for the usual VMs and VPNs, there is a need to create and route test orders, ensure all modalities are correctly set up, and HL7 is set up to handle the expedited workflow.
- Protected Time: Do all the necessary skills sets have the ability to apply time, from setting up VPN access to integrating a button in Epic?
If the team lacks experience or time to manage these clinical nuances, the project slows and may ultimately fail to realize its intended vision. This underscores why successful DICOM SR projects require close clinical leadership alongside IT, rather than being treated as purely technical integrations.
What distinguishes successful DICOM SR implementations?
Successful implementations use:
- Vendor-specific mapping of DICOM SR
- Tools with logic engines for conditional reporting, prior measurements, and calculators
- Templates that adapt to exam variability
- Processes to update for changes
- Project teams with protected time and the right skill sets
By using these strategies, DICOM SR becomes more than a means of transferring measurements—it becomes a powerful reporting automation tool that can generate preliminary reports.
Is a DICOM SR automation project worth the investment?
Yes. DICOM SR automation projects deliver a reliable return on investment by saving time, reducing errors, and enabling radiologists to focus on patient care. These projects typically require far less capital than new imaging equipment while delivering durable, repeatable operational savings over time.
Reference
This article is adapted from my original piece, first published by SourceForge [Link], republished here with permission. This version is expanded for radiology IT and clinical leaders with fewer technical details.
Andy Milkowski
Andy Milkowski