SUV Harmonization & EARL

Quick Answer

SUV (standardized uptake value = tissue activity concentration ÷ injected activity/body weight) is only meaningful compared like-with-like, yet it varies with scanner, reconstruction, uptake time, glucose, and calibration. Harmonization — most formally through EARL (EANM Research Ltd) accreditation — standardizes acquisition and reconstruction so SUV is comparable across scanners, sites, and serial studies. This underpins response assessment (PERCIST), multicenter trials, and any quantitative/AI use of SUV. The operational rule: never attribute an SUV change to biology when it may be a reconstruction or protocol difference.

Why SUV Drifts

• Reconstruction — algorithm, iterations/subsets, post-filter, resolution recovery, and point-spread-function (PSF) modeling all change SUVmax substantially (sharper reconstructions raise SUVmax).
• Uptake time — FDG SUV rises over ~60–90 min; a 10-min timing difference shifts values.
• Blood glucose and injected-activity/weight accuracy (extravasation, residual in syringe, clock errors).
• Scanner cross-calibration with the dose calibrator must be current.
• Partial-volume effect underestimates small-lesion SUV; SUVpeak (a fixed ~1-cm³ volume) is more reproducible than the single-voxel SUVmax.

The Metrics

Harmonization in Practice

• EARL defines phantom-based (NEMA image-quality) performance limits and reconstruction settings; accredited sites produce comparable SUVs. Many sites run a harmonized (EARL-compliant) reconstruction alongside their sharpest clinical reconstruction — the sharp one for detection, the harmonized one for quantification.
• For response assessment, hold technique constant across a patient's serial scans — ideally the same scanner and protocol — and use a consistent reference region (liver for PERCIST, with blood pool as backup).
• Newer PSF/ultra-high-resolution reconstructions raise SUV, so a harmonized series is used specifically for quantitative comparison; EARL2 standards accommodate PSF-era scanners.

Worked Point

A lesion reads SUVmax 8.0 on an old OSEM reconstruction and 11.0 on a new PSF reconstruction of the same acquisition. That +38% is pure reconstruction, not tumor progression — comparing them as if equivalent would falsely call progressive disease. Compare the harmonized-to-harmonized series instead.

Reporting Checklist

• State the reconstruction used (and whether a harmonized reconstruction was applied for quantitation).
• Report the metric (SUVmax vs SUVpeak/SUL) and the reference region.
• For serial studies, confirm same scanner/protocol/uptake time; flag any change.

Common Pitfalls

• Attributing an SUV change to biology when it reflects a reconstruction/protocol difference.
• Comparing a PSF/sharp reconstruction to an older harmonized one.
• Ignoring partial-volume underestimation in small lesions.
• Uncorrected extravasation or timing errors inflating apparent change.

Board Pearls

SUV is only meaningful compared like-with-like, yet it drifts with scanner, reconstruction (PSF/iterations/filter), uptake time, glucose, and calibration. Harmonization — formally via EARL accreditation — standardizes acquisition/reconstruction so SUV is comparable across scanners, sites, and serial studies (many sites run a harmonized reconstruction alongside their sharpest clinical one). For response assessment (PERCIST), hold technique constant and use a consistent liver reference.

Sharper reconstructions raise SUVmax — a PSF series can read 30–40% higher than OSEM on the same data, mimicking progression. SUVpeak/SUL are more reproducible than the noisy single-voxel SUVmax, and PERCIST uses SULpeak normalized to lean body mass with a liver reference.

Keep the drift sources straight: reconstruction and uptake time dominate, partial-volume underestimates small lesions, and cross-calibration/extravasation silently bias values. Never attribute an SUV change to biology when it may be a protocol difference — the whole point of EARL is to remove that ambiguity for multicenter and longitudinal work.

Related Pages

• Tracer: FDG; reporting: oncology response criteria (RECIST/PERCIST); physics: PET/SPECT performance & reconstruction.

Figure / Diagram Suggestions

• A same-data, two-reconstruction SUV comparison (OSEM vs PSF) showing the artifactual rise.
• A drift-source fishbone (reconstruction, timing, glucose, calibration, partial volume).

Self-Check