PSMA PET has changed the staging landscape for prostate cancer substantially over the past several years. What it has not automatically changed is the quality of quantitative data available for dosimetry planning, because staging PET and dosimetry-grade PET are different acquisition products that share the same scanner hardware but not the same requirements. This distinction matters for anyone building a Lu-177 PSMA therapy program that integrates pre-treatment imaging: the accuracy of the downstream dose calculation is bounded by the accuracy of the upstream quantification step.
Why Staging PET and Dosimetry PET Are Not the Same Product
Staging PSMA PET asks a binary question at each lesion: is this node or bone site PSMA-avid, and if so, how does its uptake compare to background? That question can be answered reliably with the standardized acquisition parameters used in clinical registration trials, including the protocols validated for 18F-DCFPyL and 68Ga-PSMA-11 in their respective regulatory submissions.
Dosimetry-grade PET asks a quantitative question: how many radioactive disintegrations occur per unit mass in this volume over time? That number feeds directly into the time-integrated activity coefficient (TIAC) calculation, which determines the absorbed dose estimate. An error of 20 percent in lesion SUVmean translates to a 20 percent error in the estimated dose to that lesion, which propagates into the PK model output and ultimately into the cycle activity recommendation. The two uses of PSMA PET are compatible in a single acquisition, but the dosimetry use imposes stricter requirements on reconstruction and quality assurance that staging use may or may not satisfy by default.
The SUV: What It Measures and What Degrades It
The standardized uptake value normalizes measured PET activity concentration in a voxel or region of interest by administered activity and body weight. SUVmax is the maximum voxel value in a defined region; SUVmean is the volume-weighted average over a segmented lesion; SUVpeak is the mean of the hottest 1 cm3 sphere within the lesion. Each metric has different sensitivity to the partial volume effect, which is the largest source of quantitative error in clinical PSMA PET.
When a tumor lesion is smaller than approximately two to three times the scanner's spatial resolution (roughly 4 to 6 mm FWHM for a modern digital PET system), activity from within the lesion is blurred outward over a larger apparent volume, and the measured concentration is lower than the true concentration. For bone metastases under 1.5 cm, this underestimation can reach 50 percent. Recovery coefficient correction, which applies a lesion-size-dependent scaling factor derived from phantom measurements, addresses the partial volume effect but requires accurate lesion volume from the co-registered CT or MRI. The correction is more reliable for spherical lesions than for infiltrative or irregular deposits.
SUVmax is most resistant to volume segmentation errors but most susceptible to noise, particularly in small lesions. SUVmean requires a well-defined segmentation boundary but is more reproducible across readers when that boundary is standardized. For dosimetry inputs, where volume as well as concentration matters, SUVmean over a consistently defined segmentation volume is the preferred metric.
Reconstruction Parameters and Their Effect on Quantitative Accuracy
PSMA PET reconstruction parameters, primarily the number of OSEM iterations and the post-reconstruction Gaussian filter width, determine the trade-off between noise suppression and spatial resolution. High iteration counts with minimal filtering maximize spatial resolution and improve partial volume recovery, but increase image noise, which degrades SUV precision in small volumes. Low iteration counts with aggressive smoothing suppress noise but systematically underestimate uptake in small lesions.
The EANM/EARL standardization framework provides reconstruction criteria designed to harmonize SUV measurements across different scanner platforms, primarily by specifying a maximum recovery coefficient range for standardized phantom spheres. EARL-compliant reconstruction is not universally adopted in clinical practice, and departments that acquired PSMA PET before optimizing for dosimetry use may have images that do not meet EARL criteria for quantitative work.
In our experience with referring images from partner sites, the most common issue is insufficient iteration count combined with aggressive smoothing, which systematically underestimates uptake in small lesions. This is not a random error: it biases the lesion dose estimate consistently downward and may create a false impression that a lesion falls below the therapeutic dose threshold when it is actually close to or within the intended dose range.
Scan Timing and Its Effect on Quantitative Reproducibility
For 68Ga-PSMA-11, optimal scan timing for both staging and dosimetry is typically 60 to 90 minutes post-injection, when tumor-to-background ratios are near their peak. For 18F-DCFPyL, 1 to 2 hours post-injection is the conventional window. Neither timing is optimized specifically for dosimetry input; both represent pragmatic compromises between image quality and department throughput.
The dosimetric consequence of scan timing variation is straightforward: the measured activity at the scan time point is one sample on the PSMA uptake time course. Two patients scanned at 60 and 90 minutes respectively will show different absolute SUV values even if their underlying uptake kinetics are identical, because they are sampling different points on the same curve. When this scan is used as the initial data point for a Lu-177 dose estimate, the timing uncertainty propagates into the TIAC calculation. Standardizing scan timing within a theranostics program to within 15 minutes of the target post-injection window is achievable with workflow coordination and materially reduces this source of variance without any change in imaging equipment or patient preparation.
From PET to Dosimetry Input: What the Downstream Pipeline Actually Needs
The YSOTOPE platform ingests three quantities from the pre-treatment PSMA PET for each lesion and organ of interest: the activity concentration at the scan time point (in kBq/ml), the segmented volume (in ml), and the exact scan time post-injection (in minutes). From these, combined with the Lu-177 SPECT series acquired after therapy initiation, the platform builds each lesion's time-activity curve and fits the PK model to generate absorbed dose estimates.
The PET data sets the initial condition for each lesion's dosimetric trajectory. An error in that initial condition propagates forward unless overridden by strong signal from the early post-therapy SPECT acquisitions. This is why acquisition protocol quality at the PET stage matters even though the SPECT series carries more weight in the final TIAC estimate. A PET that underestimates initial lesion uptake by 25 percent biases the PK model toward lower predicted tumor dose across the entire treatment course.
To be direct about the practical boundary here: a PSMA PET that meets diagnostic reporting standards is not automatically suitable as dosimetric input. The two purposes coexist in the same scanner room but require different quality assurance processes. For programs planning Lu-177 PSMA therapy with patient-specific dosimetry, implementing a PSMA PET protocol verified for SUV quantification against EARL criteria is a one-time protocol development effort, not a per-patient burden. Once validated, the same acquisition serves both staging and dosimetry purposes without additional patient contact time. The upfront investment in protocol standardization is substantially smaller than the downstream value of accurate dose calculations across a treatment course.