Dynamic PET and Tracer Kinetic Modeling

Introduction

A standardized uptake value is one number, measured at one moment, and it quietly assumes that the tracer has stopped moving. Dynamic PET drops that assumption: it follows the tracer through blood and tissue over time, and tracer kinetic modeling turns that time course into physiology. Where SUV reports how much tracer is present, kinetic modeling can separate how fast the tracer is delivered from how fast it is trapped or bound — two very different biological questions that a single late image blends together.¹⁷

For most of PET's clinical history, dynamic imaging was a research tool: it demanded a long acquisition over a single organ, arterial blood sampling, and specialized analysis. That calculus is changing. Long-axial-field-of-view (LAFOV) and total-body PET scanners can now image the entire body dynamically in one bed position, extract an input function non-invasively from a large vessel, and support shorter, lower-dose protocols. Whole-body parametric imaging — a Ki map instead of an SUV image — is moving from the research bench toward the clinic.³⁵

This article explains what dynamic PET measures, how compartment models and input functions work, how Patlak and Logan graphical analysis produce robust parameters, where the clinical value lies, and how DRPS supports quantitative PET programs through PET/CT and nuclear medicine physics and accreditation support across Florida, Maryland, Virginia, Washington DC, California, and Nevada.

Topic Explanation

From a snapshot to a time course

A conventional whole-body FDG PET is a static acquisition performed at a fixed uptake time, typically around 60 minutes after injection. It yields one activity concentration per voxel, from which SUV is computed:

SUV = \frac{C _{tissue}}{A _{injected} / BW}

where (C_{\text{tissue}}) is the measured activity concentration, (A_{\text{injected}}) is the injected activity (decay-corrected), and (\text{BW}) is body weight (or lean body mass for SUL). SUV is simple, reproducible when protocols are standardized, and clinically powerful — but it is a semi-quantitative ratio, not a physiological rate, and it is sensitive to uptake time, blood glucose, body habitus, partial-volume effects, and scan timing.⁷⁸ Our companion articles on PET SUV quantification and EARL SUV harmonization cover those confounders in depth.

Dynamic PET replaces the single snapshot with a movie. The scanner acquires a sequence of frames — short frames early, when the blood signal changes rapidly, and longer frames later — building a time-activity curve (TAC) for every region or voxel. The shape of that curve, not just its final height, carries the physiology.

What the time course reveals

Two tissues can reach the same SUV by very different routes. One may be highly perfused with modest trapping; another may be poorly perfused but avidly trapping. Their late-frame concentrations can coincide even though their biology diverges. The dynamic curve distinguishes them: the early phase is dominated by delivery and blood volume, while the later slope reflects net irreversible uptake or reversible binding. Kinetic modeling formalizes this separation, and it is the reason dynamic PET can be more specific than a single SUV in the right setting.¹⁶

Key Technical Principles

Compartment models

The workhorse of kinetic modeling is the compartment model, which represents tracer as moving between a small number of well-mixed pools with first-order rate constants. For FDG, the standard is the two-tissue compartment model: plasma, a free/unbound tissue compartment ((C_1)), and a metabolized/trapped compartment ((C_2)):

\frac{d C _{1}}{d t} = K_{1} C_{p} - (k_{2} + k_{3}) C_{1} + k_{4} C_{2}

\frac{d C _{2}}{d t} = k_{3} C_{1} - k_{4} C_{2}

Here (C_p) is the plasma (input) concentration, (K_1) and (k_2) describe transport between plasma and tissue, (k_3) describes phosphorylation (trapping), and (k_4) describes dephosphorylation. For FDG over typical scan durations, (k_4 \approx 0) (trapping is effectively irreversible), which is what makes graphical analysis so clean.⁶

The clinically central parameter is the net influx rate:

K_{i} = \frac{K _{1} k _{3}}{k _{2} + k _{3}}

For glucose metabolism, (K_i) is converted to a metabolic rate using plasma glucose and a lumped constant (LC):

MR_{glc} = \frac{C _{p, glucose}}{L C} K_{i}

The table below situates the common analysis choices.

Method	What it estimates	Input needed	Reversible or irreversible tracer	Practical note
SUV (static)	Semi-quantitative uptake ratio	None (injected dose, weight)	Either	Fast and standard; not a physiological rate
Two-tissue compartment fit	K1, k2, k3, (k4), Ki	Full input function	Either	Most complete; noise-sensitive per voxel
Patlak graphical	Net influx rate Ki	Input function	Irreversible (e.g., FDG)	Robust, linear, ideal for parametric maps
Logan graphical	Total distribution volume VT	Input function	Reversible (e.g., many neuroreceptor tracers)	Robust for binding studies

The input function

Every model needs the input function — the arterial plasma concentration (C_p(t)) that drives uptake. There are three routes:

Arterial blood sampling is the reference standard but is invasive and labor-intensive, limiting routine use.
Image-derived input function (IDIF) extracts (C_p(t)) from a large blood pool (the left ventricle or the aorta) directly in the dynamic images. LAFOV scanners make this especially attractive because the aorta and target organs are imaged simultaneously without bed motion.³⁵
Population-based input function (PBIF) uses a standardized curve shape scaled to the individual with a few late blood measurements or an image tail, enabling shorter acquisitions. One LAFOV study showed that a scaled PBIF with only 20 minutes of PET data achieved less than 15% precision error in Ki estimates compared with a full IDIF.⁵

Graphical analysis: Patlak and Logan

Fitting four rate constants per voxel is noisy, so graphical linearizations are used for parametric imaging. For an irreversibly trapped tracer, Patlak analysis transforms the data so that, after an equilibration time, a plot becomes a straight line:

\frac{C _{T} ( t )}{C _{p} ( t )} = K_{i} \frac{\int _{0}^{t} C _{p} ( τ ) d τ}{C _{p} ( t )} + V_{0}

The slope of the linear portion is the net influx rate (K_i), and the intercept (V_0) reflects the reversible plus vascular distribution volume.¹² Because the relationship is linear, (K_i) can be computed voxel-by-voxel to produce a parametric Ki image — a whole map of net influx rather than a single region value.

For reversible tracers, Logan analysis produces a plot whose late slope estimates the total distribution volume (V_T), the standard outcome for many neuroreceptor and binding studies. Compartmental fitting, Patlak, and Logan are complementary: the compartment model is the most complete description, while the graphical methods are the robust, noise-tolerant tools that make voxel-wise parametric imaging feasible.⁴⁶

A worked example

Suppose a tumor region-of-interest, after the Patlak equilibration time, gives two points on the linearized plot. The x-axis ("Patlak time," the normalized integral of the input) moves from 15 to 35 (min), and the y-axis (tissue-to-plasma ratio) moves from 3.0 to 4.2. The slope is the net influx rate:

K_{i} = \frac{4.2 - 3.0}{35 - 15} = \frac{1.2}{20} = 0.060 mL \cdot min^{- 1} \cdot mL^{- 1}

That single number captures irreversible trapping directly, independent of the reversible and vascular signal folded into a late SUV. Comparing this (K_i) before and during therapy has been shown to separate responders more distinctly than the corresponding change in SUV in some tumor settings.⁶

Clinical Impact

Oncology: response assessment and specificity

The most immediate clinical target is tumor response. Because (K_i) isolates net metabolic trapping, changes in (K_i) during chemoradiation can be a cleaner signal of response than changes in SUV, which mix perfusion, blood volume, and uptake time. Studies deriving parametric (K_i) images report response signals that are statistically more pronounced than the corresponding SUV change in selected lesions.⁶⁷ Parametric imaging also improves lesion-to-background contrast in some low-uptake settings, because the reversible/vascular background is suppressed. For established response frameworks based on SUV, see PERCIST tumor response quantification.

Cardiac and neurological quantification

Kinetic modeling is already standard in two domains. In cardiac PET, absolute myocardial blood flow and flow reserve are computed from dynamic first-pass data using compartmental kinetics, adding prognostic information beyond relative perfusion — the subject of our quantitative myocardial blood flow article. In neurology, reversible neuroreceptor and neuroinflammation tracers are routinely quantified with compartmental and Logan analysis to estimate binding and distribution volume; one first-in-human neuroinflammation study, for example, selected a reversible two-tissue model and validated distribution-volume stability to shorten scan time.⁴

Total-body dynamic imaging

The arrival of total-body PET is the biggest practical shift. A single-position dynamic acquisition can capture the input function and every organ simultaneously, enabling multi-organ kinetic analysis and whole-body parametric maps that were impossible when only one organ could be imaged at a time. The sensitivity gain of total-body systems — on the order of a 15- to 68-fold increase in effective signal collection for a scanner with a very long axial field of view — is what makes low-dose, short, or delayed dynamic protocols feasible.³ Our total-body and long-axial-field-of-view PET article covers the hardware side.

Practical Tips

1. Choose the model to fit the tracer and the question

Use Patlak analysis for irreversibly trapped tracers such as FDG when the target is net influx; use Logan for reversible binding tracers.
Do not fit four rate constants per voxel if a graphical method answers the clinical question more robustly.

2. Get the input function right

Prefer an image-derived input function from a large, well-defined blood pool; place the sampling volume carefully to minimize partial-volume and spillover error.
Consider a population-based input function with a scaling measurement to shorten the acquisition, and validate it against your own data before clinical use.⁵

3. Design the frame sequence deliberately

Use short early frames to capture the rapidly changing blood peak and longer late frames to control noise; a mismatched frame plan biases the fit.
Confirm the Patlak or Logan equilibration start time (t*) is appropriate for the tracer; starting too early violates the linearity assumption.

4. Standardize before quantifying

Kinetic parameters inherit every calibration error in the scanner. Maintain dose-calibrator, scanner cross-calibration, and clock synchronization exactly as for SUV; see PET/CT daily QC and calibration.
Keep patient preparation (fasting, glucose, uptake time) controlled; kinetic modeling reduces but does not eliminate biological variability.

Common pitfalls to avoid

Treating a Ki image like an SUV image. The units, scaling, and interpretation differ; report them distinctly.
Ignoring motion. A long dynamic acquisition is more vulnerable to patient and respiratory motion than a single static frame.
Under-sampling the blood peak. Too-long early frames blur the input function and bias every downstream parameter.
Applying a population input function without local validation. Curve shapes vary with injection protocol and physiology.

Regulatory Considerations

Quantitative PET, static or dynamic, rests on a validated, well-calibrated system and a documented protocol. The relevant frameworks include:

NEMA NU 2 — the standard for PET performance measurements (sensitivity, spatial resolution, noise-equivalent count rate, and image quality) that underpins any quantitative claim, including kinetic parameters. Current systems are characterized against the latest NEMA NU 2 edition.
EANM FDG PET/CT procedure guidelines, version 2.0 — the widely used procedural and harmonization framework emphasizing that PET is a quantitative technique requiring standardized QC/QA to make numeric values comparable across systems and sites.⁸
ACR–AAPM PET/CT accreditation and equivalent programs, which set the phantom, calibration, and QC expectations that make quantitative imaging defensible.
10 CFR Part 35 — the NRC framework governing the medical use of the PET radiopharmaceuticals themselves, including authorized use, dosage determination, and radiation safety.

Because these radiopharmaceuticals are byproduct or accelerator-produced material used under a radioactive material license, facilities must confirm whether the NRC or their Agreement State administers their program. Of the states DRPS serves, Florida, Maryland, Virginia, California, Nevada, Pennsylvania, New York, and New Jersey are Agreement States, while Washington, DC and Delaware are regulated directly by the NRC for radioactive material. Quantitative-imaging protocols do not change licensing, but they raise the importance of documented calibration and QC that an accreditation reviewer or inspector can verify.

Frequently Asked Questions (FAQs)

What is dynamic PET?

Dynamic PET is an acquisition that images the same region repeatedly from the moment of injection, producing a series of frames instead of a single static image. The result is a time-activity curve for every voxel or region, showing how tracer concentration rises and falls over time rather than only its value at one uptake time.

What is tracer kinetic modeling?

Tracer kinetic modeling is the mathematical analysis that converts PET time-activity curves into physiological parameters. Using a model of how the tracer moves between blood and tissue compartments, plus an input function describing the tracer concentration in blood, it estimates quantities such as delivery, the net influx rate, binding, and distribution volume.

How is kinetic modeling different from SUV?

SUV is a single semi-quantitative ratio from one time point, sensitive to uptake time, body habitus, blood glucose, and scan timing. Kinetic modeling uses the full time course and an input function to separate tracer delivery from trapping, yielding physiological parameters such as Ki that are less confounded, at the cost of a longer, more complex acquisition and analysis.

What is the input function and why does it matter?

The input function is the tracer concentration in arterial blood plasma over time, the driving force for uptake. Kinetic models need it to interpret tissue curves. It can be measured by arterial sampling (the reference standard), derived non-invasively from a large blood pool in the images (an image-derived input function), or approximated with a population-based curve scaled to the patient.

What are Patlak and Logan analysis?

They are graphical linearization methods. Patlak analysis applies to irreversibly trapped tracers such as FDG; a transformed plot becomes linear and its slope equals the net influx rate Ki. Logan analysis applies to reversible tracers; the slope of its transformed plot estimates the total distribution volume. Both avoid fitting every rate constant and are robust for parametric imaging.

Does dynamic PET require a long-axial-field-of-view scanner?

No, but such scanners help enormously. Conventional scanners can perform single-bed dynamic PET over one organ. Long-axial-field-of-view and total-body systems can image the whole body dynamically in one position, provide a non-invasive image-derived input function from the aorta, and enable shorter, lower-dose protocols, which makes whole-body kinetic modeling clinically practical.

Is dynamic PET worth the added complexity in routine practice?

It depends on the question. For staging, static SUV is efficient and well validated. For response assessment, cardiac flow quantification, neuroreceptor studies, and research, the added physiological specificity of kinetic parameters can justify the longer acquisition — increasingly so as total-body scanners and non-invasive input functions lower the practical cost.

Key Takeaways

SUV is a snapshot; dynamic PET is a movie. The shape of the time-activity curve, not just its final height, carries the physiology.
Kinetic modeling separates delivery from trapping. Compartment models estimate rate constants, and the net influx rate Ki isolates irreversible uptake.
The input function is essential. Arterial sampling is the reference standard, but image-derived and population-based input functions make dynamic PET far more practical.
Patlak and Logan analysis are the robust tools. Patlak's slope gives Ki for irreversible tracers; Logan's slope gives distribution volume for reversible tracers, enabling voxel-wise parametric images.
Total-body PET is the enabler. Single-position whole-body dynamic acquisition, non-invasive input functions, and large sensitivity gains move kinetic modeling toward the clinic.
Calibration underpins everything. Kinetic parameters inherit every QC error, so standardization is not optional.

Conclusion

Dynamic PET and tracer kinetic modeling answer a question SUV cannot: not merely how much tracer is present, but how the tissue handles it over time. By combining a compartment model, an input function, and robust graphical analysis, PET can report physiological parameters — net influx, metabolic rate, distribution volume, blood flow — that are less confounded than a single late uptake value.

For years this rigor was confined to research because it required arterial sampling and single-organ acquisitions. Long-axial-field-of-view and total-body scanners are dissolving those barriers, making whole-body parametric imaging and non-invasive input functions realistic. The medical physicist's role is to ensure the calibration, protocol design, and quality control that make these numbers trustworthy. As quantitative PET matures, kinetic modeling is likely to move from a specialist technique to a routine option wherever the biological question justifies looking past the snapshot.

How DRPS Can Help

Diagnostic Radiation Physics Services supports quantitative and dynamic PET programs from calibration to interpretation. This includes PET/CT and nuclear medicine physics testing and scanner cross-calibration, protocol review for dynamic and parametric acquisitions, accreditation support for ACR and equivalent programs, and medical physics consulting on quantitative-imaging methodology and QC.

DRPS supports facilities across our service locations, including Florida, Maryland, Virginia, Washington DC, California, Nevada, New York, Pennsylvania, New Jersey, and Delaware.

Quantitative PET is only as good as the physics behind it — DRPS helps make the numbers defensible.

Related Resources

References

Patlak CS, Blasberg RG, Fenstermacher JD. Graphical evaluation of blood-to-brain transfer constants from multiple-time uptake data. J Cereb Blood Flow Metab. 1983;3(1):1-7. doi:10.1038/jcbfm.1983.1. doi.org
Patlak CS, Blasberg RG. Graphical evaluation of blood-to-brain transfer constants from multiple-time uptake data. Generalizations. J Cereb Blood Flow Metab. 1985;5(4):584-590. doi:10.1038/jcbfm.1985.87. doi.org
Spencer BA, McBride K, Hunt H, Jones T, Cherry SR, Badawi RD. Practical Considerations for Total-Body PET Acquisition and Imaging. Methods Mol Biol. 2024;2729:371-389. doi:10.1007/978-1-0716-3499-8_21. doi.org
Van Weehaeghe D, Koole M, Schmidt ME, et al. [11C]JNJ54173717, a novel P2X7 receptor radioligand as marker for neuroinflammation: human biodistribution, dosimetry, brain kinetic modelling and quantification of brain P2X7 receptors. Eur J Nucl Med Mol Imaging. 2019;46(10):2051-2064. doi:10.1007/s00259-019-04369-6. doi.org
Sari H, Eriksson L, Mingels C, et al. Feasibility of using abbreviated scan protocols with population-based input functions for accurate kinetic modeling of [18F]-FDG datasets from a long axial FOV PET scanner. Eur J Nucl Med Mol Imaging. 2022;50(2):257-265. doi:10.1007/s00259-022-05983-7. doi.org
Colard E, Delcourt S, Padovani L, et al. A new methodology to derive 3D kinetic parametric FDG PET images based on a mathematical approach integrating an error model of measurement. EJNMMI Res. 2018;8(1):99. doi:10.1186/s13550-018-0454-9. doi.org
van Velden FHP, Kramer GM, Frings V, et al. Repeatability of Radiomic Features in Non-Small-Cell Lung Cancer [18F]FDG-PET/CT Studies: Impact of Reconstruction and Delineation. Mol Imaging Biol. 2016;18(5):788-795. doi:10.1007/s11307-016-0940-2. doi.org
Boellaard R, Delgado-Bolton R, Oyen WJG, et al. FDG PET/CT: EANM procedure guidelines for tumour imaging: version 2.0. Eur J Nucl Med Mol Imaging. 2015;42(2):328-354. doi:10.1007/s00259-014-2961-x. doi.org
National Electrical Manufacturers Association. NEMA NU 2: Performance Measurements of Positron Emission Tomographs (PET). Rosslyn, VA: NEMA. nema.org