Opportunistic findings on chest CT

A chest CT is ordered to answer one question. Is there a nodule, is the infection better, has the cancer spread? The radiologist answers it, and the report goes out. But the scan has also imaged the heart, the thoracic spine, the upper abdomen and every bit of lung, whether or not anyone asked. Some of what is visible there predicts serious disease years in advance, at no extra radiation and no extra cost to the patient. Using that information is what people mean by opportunistic imaging.

So, let's go over four findings that a routine chest CT can tell you about, with the evidence for each, and then look at how to make them routine without slowing the read. Clinical decisions about any individual patient remain with the treating doctor.

1. Coronary artery calcium on a non-gated chest CT

This is the strongest case, and it has a guideline behind it.

Coronary calcium scoring is traditionally done on a dedicated ECG-gated cardiac CT (gated meaning the images are timed to the heartbeat, so the heart is still). But calcium in the coronary arteries is often obvious on an ordinary chest CT too, including the low-dose scans used for lung screening and the plain CTs done in large numbers for cough, fever and follow-up of infection.

In 2016, the Society of Cardiovascular Computed Tomography and the Society of Thoracic Radiology published joint guidelines for coronary artery calcium scoring of noncontrast noncardiac chest CT (Hecht and colleagues, J Cardiovasc Comput Tomogr 2017). The key recommendations, as summarised by Guideline Central, are as follows.

That first line is clear. If a patient's chest CT shows coronary calcium and the report does not mention it, an opportunity has been missed. The ACR has since proposed a quality measure, ACRad 36, built on the same recommendation. It allows either a quantitative score or a visual ordinal assessment of each of the four main coronary arteries, and says recommendations for cardiovascular risk assessment should accompany any non-zero score.

In India, where many adults have never had a formal cardiovascular risk assessment, a line in a chest CT report saying "moderate coronary artery calcification, LAD predominant; suggest cardiovascular risk assessment" can matter a great deal.

2. Bone density and vertebral fractures

A chest CT images the thoracic spine and often the upper lumbar vertebrae. Two things can be read from it.

The first is fractures that nobody reports. Vertebral compression fractures are important: once trauma and tumour are excluded, they usually signal osteoporosis, and an osteoporotic fracture predicts further fractures. They are also easy to miss on axial images, because the height loss is best seen on a sagittal view that many radiologists never open on a chest CT. In a study of chest and abdominal MDCT, Bartalena and colleagues found vertebral fractures in 9.5% of 323 patients, and only 6 of the 41 fractures (14.6%) had been mentioned in the final report. The fix costs a few seconds: look at a sagittal reformat of the spine on every chest CT.

The second is bone density from attenuation. On abdominal CT, Pickhardt and colleagues (Ann Intern Med 2013) compared L1 vertebral attenuation with DXA in 1,867 adults. An L1 threshold of 160 HU or less was about 90% sensitive for osteoporosis, and 110 HU gave specificity above 90%. That study used abdominal scans, so the thresholds apply most directly to L1. When L1 is in the field of a chest CT, a quick region of interest can flag patients who should be referred for DXA. It is a prompt for testing, not a diagnosis.

3. Emphysema

Emphysema is often visible on CT in people who have never been told they have COPD. Mentioning it, with a grade, is useful.

In the COPDGene study, Lynch and colleagues (Radiology 2018) graded emphysema visually on CT. Every visual grade above trace was associated with higher mortality compared with no emphysema; trace emphysema was not significantly associated.

So a report that says "moderate centrilobular emphysema, upper lobe predominant" is not just describing the picture. It tells the clinician something about prognosis and is a reason to ask about spirometry and smoking.

4. Liver fat

The upper abdomen, including part of the liver, is usually in the field of a chest CT. On an unenhanced scan, fatty liver lowers liver attenuation.

In a study of 3,357 asymptomatic adults scanned with low-dose unenhanced CT, Boyce and colleagues (AJR 2010) used liver attenuation of 40 HU or less to define moderate to severe steatosis and found it in 6.2%. Milder criteria, using a liver-to-spleen ratio, raised the figure much higher. They also found clinical risk factors were poor at predicting who had it.

Those scans were colonography studies rather than chest CTs, and the measurement only works on non-contrast scans, because contrast changes liver attenuation. On a plain chest CT, though, an ROI in the visible liver (and spleen for comparison) takes seconds and can prompt a referral for proper evaluation.

Making it routine without slowing the read

The problem with opportunistic findings is not that radiologists do not know about them. It is that each one adds a step to every chest CT, and at volume the steps get skipped. A few things help.

1. Put them in the template. Add short lines for coronary calcium, spine and liver to the chest CT template, so they are a prompt rather than a memory test. (We discuss where templates help, and where they get in the way, in Structured reporting.)
2. Use software where it is reliable. Calcium detection and scoring is a good fit, because the target is bright, well defined and tedious to measure by hand. The software should propose and the radiologist decide, which is the principle behind AI is with the radiologist.
3. Carry the numbers as data. A per-artery score that lands in the report as a coded value, rather than retyped text, can be tracked over time (the DICOM side of this is in DICOM SR: AI findings a report can read).
4. Say what to do next. "Suggest cardiovascular risk assessment" or "suggest DXA" is more useful to a GP than a number with no context.
5. Agree it with referrers. Some referring doctors will welcome these lines; others will need to understand why they appear. A short note to your main referrers saves confusion.

The practical upshot is that a chest CT already contains more than the question it was ordered for, and the cost of reporting it is a few seconds per finding if the template prompts for it and software does the tedious measuring. SOMATIQ now runs across more than 100 imaging centres, and every plain chest CT read on it is a chance to say a little more than the question asked.

Related reading

• How we approached coronary calcium scoring — the method and the design choices in our viewer.
• Structured reporting — template lines that prompt opportunistic findings.
• Priors on screen — comparison matters for emphysema and calcium over time too.
• Trinity Reporting Engine — drafting the "what to do next" line without extra typing.

FAQ