Systematic 25% SUV Drop After PET/CT Detector Block Replacement
Your department operates a time-of-flight PET/CT that underwent replacement of one detector block and a vendor software upgrade eight weeks ago. Since then, technologists report that F-18 FDG SUVmax values appear about 25% lower than on comparable earlier studies. The scanner passes its daily QC: detector normalization and daily blank scan are within tolerance, the daily well-counter check source reads within ±5%, and the dose calibrator passes its quarterly accuracy test with Cs-137 and Co-57 within ±4%. Today a 64-year-old man with stage IIIB lung adenocarcinoma is restaged after four cycles of chemoimmunotherapy. His dominant lesion is unchanged at 31 mm, but SUVmax has fallen from 12.4 in January to 8.9. The oncologist has already dictated a note describing "metabolic response" and plans to stop maintenance therapy on that basis. You are asked to review the study before the report is released, and the patient is sitting in the clinic waiting for a same-day decision.
1. The SUV decrease is uniform across all patients, not lesion-specific. Work through the possible causes in order of likelihood and describe the measurements you would make today to distinguish them.
Read the worked answer
Start by defining the arithmetic, because the direction of the bias tells you which side of the equation is wrong. SUV = measured image concentration ÷ (administered activity ÷ body weight). A 25% fall across every patient can only come from an under-read on the scanner side (activity-concentration calibration, sensitivity, or reconstruction) or an over-read of the administered activity, since body weight is not systematically wrong.
Scanner side (most likely after a block replacement and software change):
- Check whether the upgrade silently changed reconstruction defaults - iterations, subsets, post-filter, matrix size, TOF/PSF switch, or a "SUV normalisation" flag. A smoother reconstruction lowers SUVmax for the same data by 10-30%.
- Perform an activity-concentration calibration (cross-calibration): a uniform cylinder phantom filled with a known F-18 concentration assayed in the dose calibrator, scanned with the clinical protocol. Tolerance is usually ±10%; a 25% error fails.
- Confirm the normalisation scan was repeated and the sensitivity re-measured after the block replacement.
Dose-calibrator side (less likely): accuracy QC passed, so a 25% error is improbable. Check the correct radionuclide setting, the geometry/dilution factor of the calibration vial, the residual-in-syringe correction, and whether the injection time recorded matches the actual clock.
Then re-reconstruct the patient with the January parameters before drawing any conclusion.
2. What do you tell the oncologist today, and how do you handle the comparison with the January study?
Read the worked answer
Hold the report. Tell the oncologist plainly that a systematic SUV discrepancy is under investigation, that the numerical SUV values on this study are not currently interpretable, and that no treatment decision should rest on them until the calibration is resolved.
What you can safely report today is the anatomical endpoint: the dominant lesion is stable at 31 mm. If the response assessment framework in use requires SUV/SUL change, it also requires that the scans be acquired and reconstructed identically; a 25% systematic shift is larger than any clinically meaningful threshold and would fabricate a response. State that the comparison is non-qualifying rather than "response".
Practical route: check whether the January images are stored and can be re-reconstructed on the current reconstruction defaults, and re-reconstruct today's study using the January parameters. If the two studies then agree, the discrepancy is reconstruction, not biology. If not, the calibration fix (and probably a repeat cross-calibration) has to come first.
Also flag in writing that stopping maintenance therapy on a 25% artefact has real consequences, and offer to call the oncologist the same afternoon once the phantom cross-calibration is complete. Document this conversation in the record, because a downstream clinician may later read only the dictated note.
3. Outline the QA program changes that would have caught this and that would prevent a recurrence.
Read the worked answer
The failure is one of change control, not of daily QC - daily blank scans and well-counter checks are designed to detect drift, not to validate quantitation after hardware or software changes.
Changes to put in place:
- Post-service release criteria. After any detector block replacement or retune: vendor normalisation, sensitivity measurement, tomographic resolution and uniformity, and a full phantom cross-calibration, signed off before the first patient. No clinical scanning in between.
- Scheduled cross-calibration. At acceptance, annually, and after any repair, compare a dose-calibrator-assayed F-18 cylinder against the scanner's measured activity concentration; tolerance ±10%. Record the ratio, not just pass/fail.
- Software upgrade protocol. Any version change triggers a documented comparison of a reference phantom reconstructed with old and new parameters, plus confirmation of which defaults govern clinical protocols and SUV.
- Longitudinal quantitation control chart. Scan the same stable phantom monthly and plot measured activity concentration and a reference SUVmax on a control chart so slow drift, not just a step change, is visible.
- Dose-calibrator audit. Annual linearity, geometry, and accuracy, plus a check that technologists record net injected activity (pre-injection minus residual) consistently.
- Training and a written authority to hold a study for any technologist who sees an unexpected systematic SUV shift.