Inter-scanner PET comparisons expose the limits of routine calibration
Positron emission tomography has long carried a quantitative reputation. In theory, it does more than visualize tracer uptake; it estimates regional differences in radiotracer accumulation and converts those signals into physiologic information that can guide research and clinical decision-making. The difficulty is that this promise depends on the scanner, the reconstruction pipeline, the acquisition protocol, and the tracer itself all behaving in a way that preserves measurement accuracy. As PET technology becomes more sophisticated, that assumption deserves repeated scrutiny rather than passive trust.
A commentary in Molecular Imaging and Biology focused on a recent inter-scanner study that compared human quantitative outcomes across three different radiotracers. The central message is straightforward: phantom testing remains necessary, but it does not settle the question of how faithfully PET performs in real patients. Human data add a layer of realism that bench-style verification cannot fully capture.
One sentence captures the problem: a scanner can pass a phantom test and still produce clinically meaningful variability in people.
Human tracer data test the promise behind quantitative PET
The appeal of PET quantification lies in its ability to move beyond visual interpretation. Regional uptake can be measured, compared, and, in principle, tracked over time with enough precision to support physiologic inference. That capability has made PET indispensable in oncology, neurology, cardiology, and other fields where subtle changes in tissue behavior matter. Yet as detectors, reconstruction algorithms, time-of-flight systems, and correction methods evolve, the numerical output can shift in ways that are not obvious on a routine image read.
The inter-scanner comparison discussed in the commentary is valuable precisely because it uses human studies rather than test objects. Human biodistribution is not a static calibration target. Motion, uptake heterogeneity, body habitus, organ-specific kinetics, and differences between tracers all influence quantitative output. A phantom can model uniform activity and known geometry, but it cannot reproduce the biological variation that determines whether a PET-derived number actually means the same thing on two different systems.
This is where the commentary draws its strongest clinical point. Standard phantom exercises, including those prescribed by NEMA, are important for benchmarking scanner performance and standardizing quality control. They remain the backbone of technical validation. Yet the resulting data can give a cleaner answer than biology ever will. In human subjects, the signal is affected by metabolism, perfusion, receptor binding, clearance, and local tissue environment, all of which can amplify or dampen apparent differences between scanners.
Why newer PET systems need more than phantom reassurance
The quantitative credibility of PET depends on more than image appearance. It depends on whether the same biologic state produces the same numeric readout across platforms and time points. As systems become more complex, small changes in reconstruction settings or detector performance can alter standardized uptake values, kinetic estimates, or other quantitative endpoints. When those measurements are used to track treatment response or compare disease burden, even modest inconsistency can complicate interpretation.
The commentary emphasizes that the inter-scanner study offers a useful complement to phantom testing because it probes the real clinical environment. That approach is also more difficult to interpret. Human studies rarely provide a single, tidy explanation for variation. Differences can reflect the scanner, but they can also reflect tracer behavior, patient-specific biology, or the interaction between the two. The result is not a simple pass-fail result; it is a more realistic picture of how PET behaves once technology meets physiology.
This difficulty is not a weakness of the method. It is the point. Quantitative imaging earns its value only if the numbers remain trustworthy when the conditions become messy. PET cannot maintain a quantitative reputation on the basis of calibration alone. It must show that the translation from measured counts to biologic information survives actual clinical use.
Three tracers, one message: quantitative accuracy is not automatic
The study discussed in the commentary compared quantitative outcomes for three different radiotracers across scanners. Even without adding unreported details, the design itself makes an important methodological statement: tracer-specific behavior matters. Radiotracers differ in how they distribute, bind, and clear, so the same scanner may not handle every agent with equal precision. A platform that performs well for one tracer may behave differently for another, especially when the biological signal is subtle or when regional uptake is highly variable.
That matters in practice because quantitative PET is often used to compare values across time, sites, or imaging systems. If a change in measured uptake reflects instrumentation rather than biology, the clinical interpretation becomes unstable. Treatment response may appear more dramatic than it is, or real biologic change may be obscured by technical drift. The stakes are especially high when PET is used to guide decisions that depend on relatively small quantitative differences rather than gross visual changes.
By pointing toward inter-scanner comparison in people, the commentary argues for a more disciplined standard of validation. Phantom studies remain indispensable for technical quality assurance and harmonization. Human studies, though more complex, are needed to determine whether those technical gains translate into reliable clinical measurements. The two approaches answer related but different questions, and PET needs both if it is to remain a credible quantitative tool.
A scanner that looks stable in a phantom can still wobble in a patient.
Quantitative PET will require ongoing recalibration of confidence
The deeper lesson is not that PET quantification is failing. It is that confidence in PET quantification should be earned repeatedly, not assumed once and preserved forever. Each increase in scanner sophistication introduces new opportunities for bias, drift, or reconstruction-related distortion. Each new tracer introduces another biologic variable. Each multicenter study raises the question of whether numbers derived at one site are truly comparable to numbers derived elsewhere.
The commentary makes a careful, methodologically grounded argument: inter-scanner studies in humans are a useful complement to standard phantom tests, but their outputs are inherently harder to interpret. That complexity should not discourage their use. It should shape how investigators, imaging physicists, and clinicians read the data. Human-based validation is messier because real patients are messier, and quantitative imaging only becomes clinically meaningful when it can withstand that complexity.
For PET to keep its quantitative reputation, ongoing reassessment is not optional. It is part of the technology itself, especially as scanners, tracers, and reconstruction methods continue to evolve.