PET/CT SUV Drift After a Detector Replacement
A PET/CT scanner in a community hospital had one detector block replaced. Two weeks later, the nuclear medicine physician reports that tumor SUVs in a follow-up lymphoma study appear 20% lower than the prior scan, while background liver SUV is unchanged. The scanner's daily QC (a cylindrical phantom with a known activity concentration) shows a uniform calibration factor within 2% of the prior value. The technologist confirms the same dose (370 MBq FDG), the same uptake time (60 min), and the same reconstruction protocol were used. The physician asks the physicist whether the change is real tumor response or a scanner artifact.
1. Given that the daily phantom QC is within 2%, how can the SUVs differ by 20%? Identify the most likely causes and the tests that distinguish them.
Read the worked answer
Phantom QC measures the system's global calibration under uniform, static conditions. It does not capture all effects that influence patient SUVs:
- Detector normalization / dead-time: A replaced block may have different light output or timing characteristics. The daily cylinder phantom is low-count and uniform, so it may not stress the normalization the way a patient does. A blank scan and a normalization scan comparing the new block against neighbors should be reviewed.
- Randoms and scatter correction: If the new block changes the randoms fraction or scatter estimate, the reconstructed activity concentration can shift, especially in high-activity regions, without changing the phantom calibration.
- Reconstruction: Confirm the same ordered-subset expectation maximization parameters, point-spread-function modeling, and post-filter were used. A protocol change is a common culprit.
- Uptake and injection: Verify the actual injected activity (residual in syringe), the actual uptake time, and the patient's blood glucose. A 20% SUV change can result from a 10-15 minute uptake difference or a partial infiltration.
- Partial-volume and lesion size: A responding lesion shrinks; smaller lesions lose signal to partial volume, lowering SUV independent of biology.
Distinguishing tests: repeat the calibration with a non-uniform phantom (e.g., an anthropomorphic or NEMA IEC body phantom with hot spheres), compare normalization and blank scans before/after the repair, and review the reconstruction log for the two studies.
2. The physicist must give the physician a defensible recommendation. What is the appropriate conclusion and what follow-up would you propose?
Read the worked answer
A 20% SUV drop with unchanged liver background is not reliably attributable to treatment response, because liver is often used as an internal reference and its stability argues against a global calibration shift but does not exclude a lesion-specific or reconstruction-related artifact.
Recommendation:
- Do not call this a response or progression on the basis of SUV alone. Interpret with anatomic imaging (CT size change), Deauville or Lugano criteria, and clinical context.
- Repeat the phantom calibration with a non-uniform phantom and review normalization/blank scans. If a defect is found, correct it and consider rescanning the patient or repeating the study.
- If no technical cause is found, compare the two studies side by side using the same reconstruction and display, and consider a follow-up scan in 6-8 weeks rather than acting on a single time point.
Documentation: Record the investigation, the phantom results, and the recommendation in the patient's record so the oncology team has a clear basis for the decision.