Skip to main content

m54 · Nuclear imaging in EP: PET and SPECT, standardized uptake and partial volume, inflammation and device infection

Warning

Education only. Not for diagnosis, treatment or any clinical decision.

In 60 seconds

Positron emission tomography (PET) and single-photon emission computed tomography (SPECT) image where a radioactive tracer goes. With the glucose analog FDG, labeled with fluorine-18, PET shows metabolically active tissue, including inflammation (Kinahan and Fletcher 2010). The standardized uptake value (SUV) divides the measured concentration by the injected activity per unit body weight, after correcting that activity for decay (Kinahan and Fletcher 2010). Clinical PET resolves only about 4–7 mm (Kinahan and Fletcher 2010), so a heart wall of similar thickness reads lower than its true uptake: the partial-volume effect (Soret et al. 2007). In cardiac sarcoidosis with ventricular tachycardia (VT), a positive baseline PET and a lack of improvement on repeat scans were associated with recurrence and events (Muser et al. 2016, 2018). For suspected infection of implanted devices, a multisociety consensus describes FDG PET/CT and leukocyte SPECT/CT findings at device pockets and leads (Bourque et al. 2024).

From the field

  • Sarcoidosis and VT ablation. In 31 patients with cardiac sarcoidosis and VT, 23 had MRI and PET before ablation: late gadolinium enhancement in 21 and abnormal FDG uptake in 15. VT-free survival was 55% at two years; enhancement, a positive baseline PET and no improvement on repeat PET were associated with recurrence (Muser et al. 2016).
  • Serial quantitative PET. In 20 patients with 92 serial scans, uptake was measured as SUVmax, a partial-volume-corrected SUVmean and lesion metabolic activity (corrected mean times volume). Patients whose lesion metabolic activity fell by less than 25% had more major cardiac events (hazard ratio 18.96, 95% CI 2.26 to 159.27) (Muser et al. 2018), a very wide interval from few patients and events.
  • Device infection. Consensus diagnostic features include intense FDG uptake at device pockets and leads that persists on images without attenuation correction (Bourque et al. 2024).
  • How reproducible is SUV? Test-retest variability around 10% was found in carefully run studies, and more in practice; inconsistent uptake times were the largest single effect in one multicenter study (Kinahan and Fletcher 2010).
  • Consensus on sarcoidosis. An HRS expert consensus statement covers the diagnosis and management of arrhythmias in cardiac sarcoidosis (Birnie et al. 2014); PubMed lists no abstract, so we cite it by title only.

Derivations

Decay and timing

A tracer’s activity falls exponentially:

\[ A(t) = A_0\,e^{-\lambda t}, \qquad \lambda = \frac{\ln 2}{T_{1/2}} \tag{1}\]

For fluorine-18, \(T_{1/2}\) = 109.73 min (Triangle Universities Nuclear Laboratory 2019); 68.5% remains an hour after injection (lab part A). Because SUV uses the injected activity corrected to the scan time, an error \(\Delta t\) in the recorded times scales SUV by \(2^{\Delta t/T_{1/2}}\): 6.5% for a 10-minute error, in line with the ~6% reported (Kinahan and Fletcher 2010).

Exact

Assumptions

  • Model: Equation 1 for fluorine-18 with the evaluated half-life (Triangle Universities Nuclear Laboratory 2019); tested against epml.nuclear.
  • Excretion and redistribution of the tracer are ignored; the curve is physical decay only. Uptake times and clock errors are set by the reader.
  • What would change the result: a different tracer, and biological changes in uptake with time, which also change SUV (Kinahan and Fletcher 2010).

Key takeaways

  • Every PET number carries a decay correction and therefore a clock.
  • Consistent timing between scans matters as much as accurate timing.

The standardized uptake value

PET measures activity concentration \(C\) (kBq/mL). To compare patients and scans it is normalized:

\[ \mathrm{SUV} = \frac{C}{A'/W} \tag{2}\]

where \(A'\) is the net injected activity decay-corrected to the scan time and \(W\) is body weight; with 1 mL of tissue taken as 1 g, SUV is dimensionless, and a tracer spread uniformly through the body would read 1 everywhere (Kinahan and Fletcher 2010). Errors in dose, residual syringe activity, weight and timing all pass straight into it (lab part B). Blood glucose and insulin also change FDG uptake (Kinahan and Fletcher 2010).

Key takeaways

  • SUV is a ratio: every input has to be right.
  • It normalizes dose and size, not biology.

Partial volume in the heart wall

Model the scanner’s blur as a Gaussian of full width at half maximum (FWHM) \(\Gamma\), standard deviation \(\sigma = \Gamma/(2\sqrt{2\ln 2})\). A wall of thickness \(w\) with uniform uptake, blurred, has the same error-function profile as a thin wall in CT (eq-m51-wall). Its maximum and its mean over the true wall recover

\[ \mathrm{RC}_{\max} = \operatorname{erf}\!\left(\frac{w}{2\sqrt2\,\sigma}\right), \qquad \mathrm{RC}_{\text{mean}} = \operatorname{erf}(b) - \frac{1 - e^{-b^2}}{b\sqrt\pi}, \quad b = \frac{w}{\sqrt2\,\sigma} \tag{3}\]

of the true uptake. With clinical resolution of 4–7 mm FWHM (Kinahan and Fletcher 2010) and an illustrative 10-mm wall, the mean recovers 86% to 76%; at 5.5 mm, 81% (lab part C). Dividing a measured mean by \(\mathrm{RC}_{\text{mean}}\) corrects it if the geometry is known; that is the idea behind partial-volume correction, which has to contend with noise and spill-in from neighboring tissue (Soret et al. 2007).

Exact

Assumptions

  • Model: Equation 3 for a slab with uniform uptake in zero background and a Gaussian blur; tested against epml.nuclear.
  • The blur range includes the 4–7 mm FWHM reported for clinical scanners (Kinahan and Fletcher 2010); wall thickness and the example uptake are illustrative.
  • What it ignores: uptake in neighboring tissue (spill-in), cardiac and respiratory motion, noise, and the curvature of the real wall.

Key takeaways

  • A wall as thick as the blur loses about a third of its mean uptake; one twice as thick, about a sixth.
  • SUVmax recovers more than SUVmean but rests on one noisy voxel (Kinahan and Fletcher 2010).

Serial scans and a response rule

A response on serial PET was defined as a decrease in lesion metabolic activity of at least 25% (Muser et al. 2018). If each scan carries independent log-normal noise with coefficient of variation \(c\), the measured ratio \(R_m\) of follow-up to baseline satisfies

\[ \ln R_m \sim \mathcal N\!\left(\ln R,\ 2\ln(1 + c^2)\right), \qquad P(\text{called a response}) = \Phi\!\left(\frac{\ln 0.75 - \ln R}{\sqrt{2\ln(1 + c^2)}}\right) \tag{4}\]

where \(R\) is the true ratio and \(\Phi\) the standard normal distribution function. With the ~10% test-retest variability reported for SUV (Kinahan and Fletcher 2010), an unchanged lesion is called a response 2.1% of the time and a true 30% decrease is missed 31% of the time (lab part E).

Key takeaways

  • A threshold on a change inherits the noise of two scans.
  • Cases near the threshold are classified partly by noise.

Clinical correlate

NotePending clinical review

This section describes published findings only. It is not guidance for patient care, and it has not yet been reviewed by an EP physician.

Where nuclear imaging shows up in EP. The cited studies use FDG PET with MRI before VT ablation in cardiac sarcoidosis and repeat PET to follow inflammation (Muser et al. 2016, 2018). An expert consensus covers arrhythmias in cardiac sarcoidosis (Birnie et al. 2014). For suspected infection of pacemakers, defibrillators and their leads, a multisociety consensus describes FDG PET/CT and leukocyte SPECT/CT (Bourque et al. 2024).

What the evidence is. The sarcoidosis data are single-center cohorts of 20 and 31 patients (Muser et al. 2016, 2018); associations with outcome are not tests of a strategy. The device-infection guidance is an expert consensus built with a modified Delphi process (Bourque et al. 2024).

What a reader should ask:

  • Which number? SUVmax, SUVmean or corrected metabolic activity behave differently (Equation 3).
  • Were the scans comparable? Same scanner, uptake time and preparation (Kinahan and Fletcher 2010).
  • Is the change bigger than the noise? Compare it with test-retest variability (Equation 4).

Visual summary

Schematic

1. Decay A = A0 e^(−λt), T½ ≈ 110 min 2. SUV C / (A′ / W) dose, weight, clock all enter 3. Partial volume thin wall reads low
Figure 1: Nuclear imaging for EP in one picture: decay sets the clock, SUV normalizes dose and size, and the scanner’s blur lowers the uptake of anything about as thin as the resolution.

Assumptions

How experts think

Before trusting an uptake value or its change, ask:

  1. Were the times right and consistent? Decay correction and uptake time both move SUV (see Equation 1 and Kinahan and Fletcher (2010)).
  2. Which inputs went into SUV? Dose, residual activity, weight, glucose (Equation 2).
  3. Is the structure thicker than the blur? If not, the mean reads low (Equation 3).
  4. Is the change larger than test-retest noise? Two noisy scans make a noisier ratio (Equation 4).
  5. What design produced the association? Small cohorts give wide intervals (Muser et al. 2018).

One sentence to keep: an uptake value is a decay-corrected, dose-normalized, blurred measurement, and its change carries the noise of two scans.

Cheat sheet

Quantity Result
Decay (Equation 1) \(A = A_0e^{-\lambda t}\), \(\lambda = \ln2/T_{1/2}\); clock error \(\Delta t\) biases SUV by \(2^{\Delta t/T_{1/2}} - 1\)
SUV (Equation 2) \(C/(A'/W)\); uniform distribution gives 1
Recovery (Equation 3) max \(\operatorname{erf}(w/(2\sqrt2\sigma))\); mean \(\operatorname{erf}(b) - (1 - e^{-b^2})/(b\sqrt\pi)\), \(b = w/(\sqrt2\sigma)\)
Response rule (Equation 4) \(P = \Phi\bigl((\ln 0.75 - \ln R)/\sqrt{2\ln(1 + c^2)}\bigr)\)

Values used in this module

Value Source
Fluorine-18 half-life 109.73 min Triangle Universities Nuclear Laboratory (2019)
Clinical PET resolution 4–7 mm FWHM; test-retest ~10%; 10-min clock error ~6% Kinahan and Fletcher (2010)
Response: lesion metabolic activity decrease of at least 25% Muser et al. (2018)

Studies in this module

Study Question Design
Muser et al. (2016) Outcomes of VT ablation in cardiac sarcoidosis Single-center cohort, 31 patients
Muser et al. (2018) Serial quantitative FDG PET and events Single-center cohort, 20 patients
Bourque et al. (2024) Nuclear imaging for cardiovascular infection Multisociety expert consensus
Birnie et al. (2014) Arrhythmias in cardiac sarcoidosis Expert consensus statement (cited by title)

Interview questions

Try each question before you open its model answer. In a real interview, say which assumptions you are making.

Q1. What does an SUV of 1 mean?

Model answer

That the tissue’s concentration equals what it would be if the injected activity were spread uniformly through the body: \(C = A'/W\) (Equation 2) (Kinahan and Fletcher 2010).

Q2. A technician’s clock is 10 minutes off. What happens to SUV?

Model answer

The decay correction uses the wrong elapsed time, so SUV is off by \(2^{10/109.73} - 1 \approx\) 6.5% (Equation 1), about the 6% reported (Kinahan and Fletcher 2010).

Q3. Why does a normal-thickness heart wall read lower uptake than it has?

Model answer

Clinical PET resolves about 4–7 mm FWHM (Kinahan and Fletcher 2010); a wall of comparable thickness spreads its counts into neighbors, so its mean recovers only part of the true value (Equation 3): about 81% for an illustrative 10-mm wall at 5.5 mm.

Q4. SUVmax or SUVmean for follow-up?

Model answer

SUVmax is less affected by region drawing and partial volume of the mean, but it rests on one voxel and is noisier (Kinahan and Fletcher 2010). Partial-volume-corrected means and metabolic activity were used in a sarcoidosis cohort (Muser et al. 2018). Whatever is chosen must be measured the same way each time.

Q5. How would you correct for partial volume, and what can go wrong?

Model answer

Divide by a recovery coefficient computed from the known geometry and resolution (Equation 3), or use more complete methods (Soret et al. 2007). It needs the true thickness, ignores spill-in from neighbors in its simplest form, and amplifies noise.

Q6. A lesion’s activity fell 20% on repeat PET. Did it respond?

Model answer

Not by a 25% rule (Muser et al. 2018), and with ~10% test-retest noise per scan (Kinahan and Fletcher 2010) a change of that size is within what noise alone often produces (Equation 4). Check that uptake time, scanner and preparation were the same.

Q7. How often would noise alone call an unchanged lesion a response?

Model answer

With 10% log-normal noise per scan, \(\Phi(\ln 0.75/\sqrt{2\ln 1.01}) \approx\) 2% (Equation 4). The bigger problem is the other side: a true 30% fall is missed about 31% of the time.

Q8. The hazard ratio for non-response was 18.96. How much weight does it carry?

Model answer

Its 95% interval ran from 2.26 to 159.27, from 20 patients (Muser et al. 2018): the direction is suggestive, the size is very uncertain. It is an association in a small cohort, not an estimate to plan around.

Q9. What do PET and SPECT measure differently?

Model answer

PET detects the two photons from positron annihilation in coincidence, which defines each line of response; SPECT detects single photons and needs a collimator to know their direction. Both image tracer distribution; which tracer and which scanner depend on the question, for example FDG PET or leukocyte SPECT for infection (Bourque et al. 2024).

Q10. What does nuclear imaging add for a suspected device infection?

Model answer

It can show uptake at the pocket and along leads, and extracardiac involvement; consensus features include intense FDG uptake on device pockets and leads that persists on non-attenuation-corrected images (Bourque et al. 2024).

Q11. Name sources of error in an SUV.

Model answer

Dose and residual syringe activity, body weight, clock synchronization, uptake time, blood glucose and insulin, scanner calibration, reconstruction and smoothing, partial volume and respiratory motion (Kinahan and Fletcher 2010).

Q12. Why might cardiac FDG uptake differ between two scans with no change in disease?

Model answer

Different preparation (glucose and insulin), uptake time, scanner or reconstruction, and noise (Kinahan and Fletcher 2010); and for thin myocardium, any change in effective resolution changes the partial-volume loss (Equation 3).

Code lab

The lab goes from decay and SUV to partial volume on a synthetic heart and the statistics of a response rule.

Part What you do Data
A Decay and timing (Equation 1) published half-life
B The standardized uptake value (Equation 2) our own numbers
C Partial volume in a wall (Equation 3) published scanner resolution
D An inflamed region on a synthetic ventricle our own phantom
E Serial scans and a 25% rule (Equation 4) published noise level and rule
F Reading the evidence published results

Each part ends with checks: lines that pass with the code as given. If you change something and a check fails, that is information, not an error to hide.

Self-check

Short checks you can do in your head. Open each answer after you have tried it.

S1. What fraction of fluorine-18 remains after two half-lives?

Answer

One quarter (Equation 1).

S2. 350 MBq injected (decay-corrected), weight 70 kg, C = 10 kBq/mL. SUV?

Answer

\(10/(350/70) = 2\) (Equation 2).

S3. Does a heavier patient with the same dose and concentration get a higher or lower SUV?

Answer

Higher: SUV is proportional to weight at fixed \(C\) and \(A'\) (Equation 2).

S4. Which recovers more of the true uptake in a thin wall, the maximum or the mean?

Answer

The maximum (Equation 3).

S5. If per-scan noise doubles, what happens to the spread of the measured ratio?

Answer

It roughly doubles too, since the log ratio’s SD is \(\sqrt{2\ln(1 + c^2)} \approx \sqrt2\,c\) for small \(c\) (Equation 4).

S6. True or false: a positive baseline PET caused VT recurrence in the cited cohort.

Answer

False: it was associated with recurrence in an observational cohort (Muser et al. 2016).

Literature figures

These figures are worth looking at alongside this module. We link to them instead of copying them: some licenses do not allow reuse here, and others we could not confirm. Each entry says what to look for and where this module explains it.

  • Sources of SUV error. Kinahan and Fletcher show how recovery falls for small objects, how SUV changes with uptake time and smoothing, and calibration drift over years [Kinahan and Fletcher (2010); free full text]. This module: the standardized uptake value and partial volume.

Further reading

References

Birnie, David H., William H. Sauer, Frank Bogun, et al. 2014. “HRS Expert Consensus Statement on the Diagnosis and Management of Arrhythmias Associated with Cardiac Sarcoidosis.” Heart Rhythm 11 (7): 1305–23. https://doi.org/10.1016/j.hrthm.2014.03.043.
Bourque, Jamieson M., Ulrika Birgersdotter-Green, Paco E. Bravo, et al. 2024. “\(^{18}\)F-FDG PET/CT and Radiolabeled Leukocyte SPECT/CT Imaging for the Evaluation of Cardiovascular Infection in the Multimodality Context: ASNC Imaging Indications (ASNC I\(^{2}\)) Series Expert Consensus Recommendations.” Journal of Nuclear Cardiology 34: 101786. https://doi.org/10.1016/j.nuclcard.2023.101786.
Kinahan, Paul E., and James W. Fletcher. 2010. “Positron Emission Tomography-Computed Tomography Standardized Uptake Values in Clinical Practice and Assessing Response to Therapy.” Seminars in Ultrasound, CT and MRI 31 (6): 496–505. https://doi.org/10.1053/j.sult.2010.10.001.
Muser, Daniele, Pasquale Santangeli, Simon A. Castro, et al. 2018. “Prognostic Role of Serial Quantitative Evaluation of \(^{18}\)F-Fluorodeoxyglucose Uptake by PET/CT in Patients with Cardiac Sarcoidosis Presenting with Ventricular Tachycardia.” European Journal of Nuclear Medicine and Molecular Imaging 45 (8): 1394–404. https://doi.org/10.1007/s00259-018-4001-8.
Muser, Daniele, Pasquale Santangeli, Rajeev K. Pathak, et al. 2016. “Long-Term Outcomes of Catheter Ablation of Ventricular Tachycardia in Patients with Cardiac Sarcoidosis.” Circulation: Arrhythmia and Electrophysiology 9 (8): e004333. https://doi.org/10.1161/CIRCEP.116.004333.
Soret, Marine, Stephen L. Bacharach, and Irène Buvat. 2007. “Partial-Volume Effect in PET Tumor Imaging.” Journal of Nuclear Medicine 48 (6): 932–45. https://doi.org/10.2967/jnumed.106.035774.
Triangle Universities Nuclear Laboratory. 2019. \(^{18}\)F Ground-State Decay: Evaluated Half-Life (Recommended Value, August 2019). TUNL Nuclear Data Evaluation Project. https://nucldata.tunl.duke.edu/nucldata/GroundStatedecays/18F.shtml.