The parts of a report, how radiologists express certainty, standard categories, and turning a report into plain language
For most clinicians, the report is the scan. Few will look at the images themselves; they act on what the radiologist wrote. Patients now often read their reports in an online portal before anyone explains them. This lesson walks through Mr. Diaz's CTPA report line by line. About 15-20 minutes. New terms are in the Module 2 glossary (opens in a new tab).
By the end of this lesson you can:
A radiology report is the radiologist's written interpretation of the images: what they saw, what they think it means, and what should happen next. It is a permanent part of the medical record and a legal document. It is also, as you saw in Lesson 2.3, the source of the labels for many AI datasets.
Most reports follow the same structure, whether the scan is a chest X-ray or a brain MRI. Many hospitals use structured report templates, with fixed headings and sometimes fixed choices, which make reports easier to read and easier for computers to process.
Click each part of the report to see what it is for and what to look for. The report is fictional but written the way real reports are.
Each dashed box is one standard section. Explore all nine.
The exam name, whether contrast was used, and the date and time. Check it matches what you ordered and the right patient. "PE protocol" tells you the contrast was timed for the lung arteries.
Usually copied from the order (Lesson 2.2). It tells you what the radiologist knew when they read the scan. If the history was vague, the read may have been less focused. "Pleuritic" means the pain is worse on breathing in.
Earlier imaging is one of the radiologist's most useful tools: a nodule that has not changed in two years is much less worrying than a new one. "No prior CT" here means nobody knows yet whether the nodule is new.
Details of the protocol and contrast. Look for any statement on quality: "adequate opacification" means the arteries filled well with contrast, so a clot could be seen if present. If quality was poor, this is where it is said, and it limits what the report can rule out.
A systematic description of what the radiologist saw, including normal things and small findings that did not make the impression, like the trace effusion and the opacity at the lung base. A filling defect is a dark gap inside a contrast-filled vessel: the clot. The series and image numbers let anyone find the nodule in PACS. The right-to-left ventricle ratio is checked because a big clot can strain the right side of the heart.
The radiologist's conclusion, numbered, with the most important finding first. It answers the clinical question ("yes, PE") and lists anything else that needs attention. Many people read only this part. See the next section for why that is risky.
Suggested next steps, often based on published guidelines. The Fleischner Society guidelines say how to follow lung nodules found by chance, based on their size, type and the patient's risk. This line is the one most easily lost after discharge (step 9 in Lesson 2.2).
For critical results, the report documents the direct call: who was told, at what time, and that they repeated it back ("read-back") to confirm. This protects the patient and is a legal record that the finding was passed on.
The signing radiologist takes responsibility for the report. Watch for "preliminary" versus "final": overnight preliminary reads can change when finalized, and an addendum is added if something is corrected later.
Reading the impression first is sensible: it gives you the answer in a few lines. But stopping there can miss things. In Mr. Diaz's report, the findings section mentions a small opacity at the lung base and a trace effusion. Neither made the impression, because the radiologist judged them less important, but they help explain his pain when he breathes. A clinician who knows the patient may find a "minor" finding matters a lot.
Speech recognition also makes mistakes. "No evidence of pneumothorax" can become "evidence of pneumothorax," and "left" can become "right." If something in the impression does not fit the findings, or the patient, phone the radiologist. They would much rather you ask.
An AI tool that summarizes reports in one line would almost certainly keep the PE and might drop the nodule. What would be lost? Who would notice?
Images rarely give a yes-or-no answer. Radiologists express how confident they are with particular phrases, sometimes called hedging language. There is no single agreed scale, and different readers interpret the same phrase differently, which is a real source of confusion. Some departments now publish a lexicon (a fixed list of phrases with approximate meanings).
Drag each phrase into the box for how likely the radiologist thinks the finding is. Approximate meanings in common use; your hospital may differ.
Hedged phrases are hard for AI too. When a program turns reports into training labels (Lesson 2.3), "cannot be excluded" must become yes, no or "uncertain," and different choices make different models.
For some common problems, radiologists use a scored category so everyone reads the same result the same way. The American College of Radiology publishes a family of these, called Reporting and Data Systems (RADS). Each pairs a number with a risk level and a recommended next step.
| System | Used for | What the number tells you |
|---|---|---|
| BI-RADS | Breast imaging (mammography, ultrasound, MRI) | 0 (needs more imaging) to 6 (known cancer). 4 and 5 usually mean biopsy. |
| Lung-RADS | Lung cancer screening CT | How suspicious nodules are, and when to scan again. |
| LI-RADS | Liver lesions in people at risk of liver cancer | How likely a liver lesion is to be cancer. |
| TI-RADS | Thyroid nodules on ultrasound | Whether a nodule needs a needle biopsy or follow-up. |
| PI-RADS | Prostate MRI | How likely a significant prostate cancer is. |
| Fleischner | Lung nodules found by chance (not part of RADS) | Follow-up timing based on size, type and risk. Used for Mr. Diaz's nodule. |
You do not need to memorize the categories. Learn to spot them in a report and look up what the number means. They are also a natural target for AI, because the output is a clear category.
Mr. Diaz has read his report in the patient portal and is worried. He asks his care team what it means. Imagine you are the one answering. Explaining a report in plain words is a skill you will use for your whole career.
One way to say it: "The scan found a small blood clot in one of the blood vessels in your right lung. That explains your breathlessness and the pain when you breathe. It is not putting strain on your heart, and the blood thinners we started will treat it. The scan also showed a small spot in your lung, about the size of a pea. Most spots like this are harmless, but we will do another CT scan in 6 to 12 months to make sure it has not changed."
Things to check in yours:
Reports connect to AI in three ways you have now seen:
Module 3 looks at how such tools are tested and used in practice.
Answer each question to continue. Any answer is fine; read the explanation.
Which part of the report gives the radiologist's overall conclusion?
The impression is the bottom line, ordered by importance. The findings describe everything seen, including smaller things that may not appear in the impression.
A report says "small pneumonia cannot be excluded." What does this most likely mean?
"Cannot be excluded" signals uncertainty. It often leans toward unlikely but leaves the decision to the clinical picture. If it matters, call the radiologist.
You only read the impression of Mr. Diaz's report. Which of these would you miss?
Those findings are only in the findings section. They are minor, but they help explain his pain. Read the whole report.
You have finished Module 2. You now know what radiology is, what an image is to a computer, how a scan moves through the hospital, and how to read the result. Module 3 asks how AI tools that work on all of this are judged and used.