Structured reporting in radiology
By SOMATIQ team ·
Medically reviewed by Dr. Vishwanath R S, MD, CIIP
Every radiologist has an opinion on this one, and most of them are right about something. The free-text camp says a report is a consultation, not a form. A good radiologist writes what matters for this patient, in the order that matters, and a template forces them to fill boxes that do not. The structured camp says referring doctors cannot find the answer in a paragraph, that important negatives get left out, and that half of what passes for "style" is really inconsistency.
Our view is that both are describing real problems, and the answer is not one or the other. It is knowing where structure earns its place. So, let's first separate what people mean by "structured", then look at where good templates come from, what the evidence says, how this plays out in Indian practice, and where AI changes the old trade-off.
What "structured" actually means
The word gets used for three different things, so it helps to separate them.
Most of the argument is really about level 2. Level 1 is uncontroversial. Level 3 is where things are slowly heading, especially for AI, because it is the level at which an AI measurement can flow into a report as data rather than being retyped. (On the imaging side, the DICOM object that carries coded AI findings is the Structured Report; see DICOM SR: AI findings a report can read, and for the FHIR route, FHIR for AI results.)
Where the templates come from
You do not have to write templates from scratch. RSNA runs RadReport, a free library of reporting templates at radreport.org. Many were reviewed by an international panel, others were contributed by RSNA members and the European Society of Radiology, and many use RadLex terms and RadElement common data elements. That last part matters for level 3: a template built on common data elements produces reports that software can parse.
For specific diseases, the ACR's Reporting and Data Systems go further. The ACR describes them as standard terminology and assessment categories for clear, consistent, actionable reports. The ones most Indian radiologists use every week:
| System | Used for | What it gives the referring doctor |
|---|---|---|
| BI-RADS | Breast imaging | An assessment category tied to a management recommendation |
| Lung-RADS | Lung cancer screening CT | A nodule category with a follow-up interval |
| TI-RADS | Thyroid ultrasound | Points for nodule features and a rule on when to do FNAC |
| PI-RADS | Prostate MRI | A likelihood score for clinically significant cancer |
The ACR list also includes LI-RADS, O-RADS, C-RADS, NI-RADS, VI-RADS and others. The strength of these systems is not the template. It is that a category means the same thing to the radiologist who wrote it and the surgeon who reads it.
What the evidence says
There is decent evidence that structure helps the person reading the report.
Schwartz and colleagues, Radiology, 2011. Radiologists, surgeons and oncologists rated 330 body CT reports.
Brook and colleagues, Radiology, pancreatic cancer CT
Surgeons judged a much larger share of the structured reports to have enough information for surgical planning.
The second study is the more persuasive one for us. Pancreatic staging has a defined list of things the surgeon needs: vessel involvement, degree of contact, variant anatomy. Free text, written at speed, skips some of them; a template does not forget.
The evidence is weaker on whether structure improves diagnostic accuracy itself. Structure mostly makes sure the right things get said and are easy to find, which is valuable, but it is a different claim.
Indian practice realities
Here is where theory meets a Tuesday afternoon at a busy centre. Five things tend to decide whether templates survive.
Volume. A radiologist reading a large daily load across CT, MRI and ultrasound does not have time to click through forty dropdowns per case. Templates that slow the read get abandoned quietly, and you are back to Word files.
Copy-paste templates. Many centres already use "templates": a normal report in a Word file that gets copied and edited. This is the worst of both worlds. The pre-filled "normal" text survives into reports where it is not true, and nobody notices because it reads fluently. A left kidney described as normal in a nephrectomy patient is a familiar example.
Referring doctor expectations. A clinician who has received a certain style from a centre for ten years may not welcome a sudden change. Introduce templates gradually, starting with studies where the benefit is obvious.
Typing and dictation. Many radiologists still type, or dictate to a transcriptionist who types. Structure that depends on rigid fields fights both habits.
Mixed referral base. A centre's referrers range from GPs to oncologists at a tertiary hospital. The GP wants a clear impression; the oncologist wants measurements against the last scan. One report has to serve both.
A practical middle path
What we would suggest to a centre owner or a lead radiologist:
Use RADS wherever one applies, within the population it was built for. Mammography, thyroid ultrasound, prostate MRI, lung cancer screening CT. These systems exist so that a category means the same thing everywhere, and referring doctors increasingly expect it.
Template the staging and protocol studies. Pancreatic, rectal and cervical cancer staging, liver MRI, prostate MRI: places where the referring team needs a checklist answered.
Keep free text for the impression. The impression is where the radiologist's judgement lives. Do not turn it into a dropdown.
Never pre-fill "normal" without review. If a template has defaults, the radiologist should have to confirm each section, or at least the system should flag sections that were never touched.
Review templates every year. Guidelines change, and a TI-RADS template built five years ago may not match the current version.
Where AI helps
This is the part that has changed in the last two years. The old trade-off was that structure cost the radiologist time. AI can remove most of that cost.
A radiologist dictates naturally, the way they always have. Software turns it into a structured report: findings organised under the right headings, measurements captured, RADS categories suggested from the described features, and an impression drafted for the radiologist to edit. The radiologist keeps their voice; the referring doctor gets the structure. Voice is already how many of our radiologists move through their day (see Talking to the worklist), so dictating into a structure is a small step.
That is the idea behind the Trinity Reporting Engine by SOMATIQ. Used on its own, it runs in the browser beside whatever PACS the centre already uses, structures the radiologist's observations, helps refine the impression, and produces a finished report that can be pasted into any RIS. Its first version had more than 100 sign-ups at IRIA 2026 (we wrote about that week in What radiologists told us at IRIA 2026), which told us the demand is real. Inside the full SOMATIQ platform, reporting also has built-in speech-to-text, LLM-assisted drafting with impression checks and inconsistency detection, and live co-reporting between radiologists and transcriptionists, which keeps the transcriptionist workflow many Indian centres rely on rather than fighting it.
AI should not decide the category on its own, and it should never invent findings the radiologist did not describe. The radiologist signs the report, so the radiologist stays in charge of what it says.
So where does that leave a centre deciding how far to go with templates? The practical upshot is to use structure where it has a clear reader (RADS, staging, protocol studies), keep the impression in the radiologist's own words, and let software carry the cost of structure rather than the radiologist. Done that way, the free-text and structured camps both get what they were asking for.
Related reading
- DICOM SR: AI findings a report can read — how coded AI findings reach the report.
- FHIR for AI results — mapping AI measurements into a reporting system.
FAQ
What is structured reporting in radiology?
It is reporting in a consistent, predefined format, from a standard layout up to templates with defined fields and coded values, so the referring doctor can find key information and software can read it.
Are structured reports better than free-text reports?
Studies have found referring doctors rate structured reports higher for content and clarity, and that they include more of the key features needed for decisions such as cancer surgery planning. Evidence on diagnostic accuracy itself is less clear.
What are RADS in radiology?
Reporting and Data Systems, such as BI-RADS, Lung-RADS, TI-RADS and PI-RADS, are standard vocabularies and assessment categories for specific organs and patient groups; some link each category to a recommended next step, while others mainly standardise the risk assessment.
Where can I find free radiology report templates?
RSNA's RadReport library at radreport.org offers free templates, many reviewed by an international panel and built on standard RadLex terms.
Ready to see your workflow on one platform?
We will walk you through the whole workflow, from registration to signed report.