Standard uptake values (SUV; also written standardised or standardized uptake value) are the PET number: how much radiotracer piled up in a voxel, normalized so two scans can be compared. It is not a Hounsfield unit. It is not a dose-volume histogram. It is not the calculator that turns injected dose and weight into SUVbw.
If you meant the SUVbw calculator → SUV calculator. If you meant Hounsfield units (CT density) → what is Hounsfield units. If you meant CT window → CT scan window settings. If you meant DVH / radiotherapy dose → dose-volume histogram. If you meant tissue perfusion → tissue perfusion. If you meant cine MRI → MRI with cine.
If the work is overlaying an SUV measurement or running a PET model inside a viewer, PYCAD is the imaging piece (viewer / model), not a nuclear-medicine vendor. The free SUVbw calculator is https://pycad.co/tools/suv-calculator/.
What the number is
A PET scanner does not see cells. It sees annihilation photons from a positron-emitting tracer — most often 18F-FDG, a glucose analog. Tissues that burn glucose (tumour, infection, brown fat, a recently used muscle) take more FDG. SUV turns that local activity into a ratio so a 90 kg patient and a 55 kg patient are not compared on raw counts.
The body-weight form, SUVbw, is:
SUV = (activity concentration in the tissue) / (injected activity / body weight)
Units have to match (typically MBq/ml over MBq/kg, which leaves g/ml). Decay-correct both the tissue measurement and the injected dose to the same clock. If the injected activity is wrong — infiltration, residual in the syringe — every SUV downstream is wrong.
That is the formula. The free calculator that runs it is the SUVbw tool, not this page.
SUVmax, SUVmean, SUVpeak
One lesion is many voxels. Reports name which summary they used.
| Metric | What it is | Why it is used | What breaks it |
|---|---|---|---|
| SUVmax | Hottest single voxel in the ROI | Easy, repeatable, the number most reports quote | Noise; one hot pixel can jump between reconstructions |
| SUVmean | Average over the whole ROI | A better picture of the whole lesion | The contour. Extra normal tissue pulls it down |
| SUVpeak | Mean inside a small fixed sphere (often ~1 cm3) on the hottest part | More stable than max; preferred in many trials | Still needs a consistent sphere and the same recon |
A lung nodule with SUVmax 8.5, SUVmean 5.2, SUVpeak 7.9 against a liver around 2–3 is a metabolic outlier. That is a map for the next step (usually biopsy), not a diagnosis by itself.
SUVlbm and SUVbsa
Fat takes almost no FDG. SUVbw divides by total body weight, so a heavier patient can look “colder” for the same tumour. Two common corrections:
- SUVlbm (lean body mass; the same idea as SUL). Height, weight, and sex go into a lean-mass formula. This is the normalization PERCIST uses.
- SUVbsa (body surface area). Same job, different size term.
Pick one normalization and keep it for the baseline and the follow-up. Mixing SUVbw on scan 1 with SUVlbm on scan 2 is not a treatment response.
Liver and blood-pool as the internal reference
There is no universal “normal SUV.” A value of 4 can be noise in the bladder and a problem in the lung. Sites use the patient’s own liver — and often mediastinal blood pool — as the yardstick on that scan.
Typical adult fasting FDG references, not cutoffs: liver SUVmean commonly sits around 2–3; mediastinal blood-pool is usually lower, often ~1.5–2. A lesion is read against those internals and the anatomy, not against a magic number copied from another hospital.
The much-quoted SUVmax > 2.5 rule of thumb (lung nodules) is a historical filter, not a cancer test. Well-differentiated tumours sit under it. Infection sits over it.
Diagnose, monitor, recurrence
Three jobs, one number:
- Diagnose / stage. A cold mass on CT is often left alone; a hot one is biopsied or staged. Nodes and distant deposits that are still small on CT can already be FDG-avid.
- Monitor treatment. Metabolic drop can show up before the long-axis shrinks. A lymphoma SUVmax falling from the teens into the liver range after two cycles is the treatment working, not a smaller blob on CT.
- Recurrence. A new hot focus on a surveillance PET can be the first sign the disease is back — still a candidate, still needing anatomy and, often, tissue.
PERCIST, if you actually need a protocol
PERCIST 1.0 (Wahl et al., Journal of Nuclear Medicine 2009) is a response rule set, not a synonym for “the SUV went down.” It uses SUL_peak, not SUVbw-max.
- Measurable target: SUL_peak at least 1.5 × mean liver SUL + 2 SD (and at least 2.0).
- Complete metabolic response: FDG avidity in the targets resolves (indistinguishable from background).
- Partial metabolic response: ≥30% drop in SUL_peak and an absolute drop of at least 0.8 units.
- Progressive metabolic disease: ≥30% rise and an absolute rise of at least 0.8, or a new FDG-avid lesion.
- Stable: neither partial response nor progression.
Same scanner class, same uptake time, same reconstruction. A 25% dip on a different recon is not PERCIST.
What throws the number
SUV is closer to a blood-pressure reading than to a height. Technical and biological knobs both move it.
- Uptake time. Most FDG protocols target ~60 minutes from injection to scan. A 60-minute baseline and a 90-minute follow-up are not a pair.
- Reconstruction. OSEM vs a regularized algorithm can move SUV by a double-digit percent on the same raw data. Lock the recipe.
- Calibration and dose. An uncalibrated gantry, a residual syringe, or an infiltrated injection scales every voxel.
- Glucose. FDG competes with glucose. Many sites will not scan if capillary glucose is high (a common institutional ceiling is ~200 mg/dL / 11.1 mmol/L).
- Fasting and insulin. A 4–6 hour fast is the usual FDG prep. A meal shoves tracer into muscle and fat and steals contrast from the lesion.
- Exercise and talking. Muscle that just worked lights up. Chewing gum is a classic false “neck node.”
- EARL / EANM accreditation exists because two scanners will not otherwise speak the same quantitative language. That is a trial / harmonization problem, not a PYCAD product.
Mimics and quiet tumours
High SUV is metabolic activity, not a histology.
| Mimic | Examples | Why it lights up |
|---|---|---|
| Infection | Pneumonia, abscess, TB | Bacteria and the immune response both take FDG |
| Inflammation | Sarcoid, arthritis, post-op bed | Macrophages and neutrophils are glucose-hungry |
| Repair | Healing fracture, biopsy track | The repair crew is metabolically loud |
| Physiology | Brown fat, strained muscle, vocal cords | Normal tissue doing work, or staying warm |
The other failure mode is a quiet cancer. Prostate adenocarcinoma, some well-differentiated lung adenocarcinomas, and many renal-cell carcinomas are poorly FDG-avid. A low SUV does not clear those. Different tracers (PSMA, amino-acid, DOTATATE) exist for a reason; they are not this article.
Where the number came from
Early PET in the 1960s–70s was a picture. Phelps, Hoffman, and Kuhl’s generation of systems (including the commercial ECAT I) plus FDG made quantification possible: attenuation correction, a known injected dose, and a ratio instead of “it looks bright.” SUV is that ratio. It is not new software, and it is not a workstation brand.
FAQ
Does a high SUV mean cancer?
No. It means high FDG (or other tracer) uptake. Infection, inflammation, brown fat, and a healing bone do the same. The report is the number plus the CT/MRI anatomy plus the history. Tissue is still the diagnosis.
What is a “normal” SUV?
There isn’t one number. Use the patient’s liver (~2–3 SUVmean on typical adult fasting FDG) and blood-pool (usually lower, ~1.5–2) as the internal reference on that scan. Fluid and necrosis sit under ~1. The 2.5 cutoff is a lung-nodule habit, not a law.
Why the 4–6 hour fast?
FDG is a sugar analog. A meal raises glucose and insulin; muscle and myocardium take the tracer; the lesion loses contrast; the SUV is no longer the SUV from the baseline. That is why the prep sheet exists.
Is this the same as Hounsfield units?
No. HU is CT attenuation. SUV is PET tracer concentration. Window/level is 3309; HU is 6213.
Does PYCAD sell a PET workstation?
No. Viewer / model if the measurement has to live inside an app. The calculator is already at /tools/suv-calculator/.
Keep Reading
Related Articles
Explore the full Quantitative Imaging HubAll services, tools, and guides on this topic — in one place.
Visit Hub →