Medical Physicist's Notes · Ge-68 Line Source

Four bands of normalization cadence — plus the drift modes between them.

PET normalization is not a once-a-year event. The full NEMA NU-2 normalization runs quarterly and after service; lighter daily / weekly checks catch detector drift before it contaminates the quarterly baseline. This post draws the four cadence bands and lists the drift modes that catch programmes between scheduled normalizations.

The four cadence bands

When to test, what to test, what crosses the action threshold

Band	Trigger	Test	Pass	Fail action
Daily / morning	Before the first clinical PET acquisition of the day	Coincidence detector counts on the Ge-68 cylindrical phantom; sinogram uniformity check	Detector counts within ±5 % of yesterday's value; sinogram visually uniform	Hold clinical scanning. Repeat after detector warm-up; check for a tripped PMT module
Weekly	One scheduled slot per week (typical Monday morning)	Full uniformity + slice-sensitivity profile from a Ge-68 cylindrical or line source	Axial uniformity < 5 %, slice-sensitivity profile within tolerance	Investigate single-slice anomalies; consult service engineer if root cause is a detector / electronics fault
Quarterly	Once per quarter; aligned with cyclotron radiopharmacy QA	Full NEMA NU-2 normalization scan with rotating Ge-68 line source; per-crystal sensitivity map regenerated	New normalization file accepted; CT/PET co-registration verified; quantitative SUV calibration on F-18 cylinder within ±10 %	Quarantine the normalization until service engineer inspects detector electronics
After service	Detector block swap, gantry mechanical service, software upgrade, scanner relocation	Full NEMA NU-2 normalization + uniformity + sensitivity + SUV-cylinder cross-check	Pre-service / post-service comparison within stated tolerance; baseline reset	Hold clinical scanning until baseline is recovered and documented in the service log

Source: AAPM TG-181 PET QA guidance; IAEA Human Health Series 27; NEMA NU-2-2018.

Drift modes

Four ways the scanner can shift between normalizations

Background sinogram counts drift up over time as the Ge-68 source decays — the activity correction has to be applied before comparing across weeks. The 271-day half-life means ~0.26 % decay per day; over a quarter that is ~24 %. The QA software should be applying this automatically, but verify the correction the first time a new source is installed.
Spatial drift (axial-uniformity asymmetry creeping in across weeks) usually points at one PMT module starting to drift its gain. Catch it on the weekly uniformity scan; do not wait for the quarterly normalization.
Energy-window drift (511 keV peak shifting) shows up as a count-rate change with no spatial pattern. Common after a power-cycle or air-conditioning event. Re-run the energy calibration before the next normalization.
After a software upgrade, ALWAYS re-run a full normalization. Software changes can move the normalization-file convention silently; the pre-upgrade normalization may load and apply but read the detector indices wrong.

Source: AAPM TG-181; manufacturer-specific service guidance.