Clinical Robustness of FDG-PET/CT Quantitative Metrics Post-Harmonization in a Multicenter, Cross-Scanner Setting

Area of Science:

• Nuclear Medicine
• Medical Imaging
• Oncology

Background:

• Quantitative metrics in FDG PET/CT vary across scanners and reconstruction methods, limiting comparability.
• Harmonization reduces inter-scanner variability in Standardized Uptake Values (SUVs), but its effect on lesion-level ranking is unclear.
• This study assesses the robustness of PET quantitative metrics and rank preservation post-harmonization.

Purpose of the Study:

• To evaluate the impact of phantom-based harmonization on the preservation of lesion-level ranking in clinical FDG PET/CT data.
• To identify the most robust quantitative PET metric for multicenter studies.

Main Methods:

• Retrospective analysis of phantom and clinical data (head and neck malignant melanoma, adenoid cystic carcinoma) from three institutions using four PET/CT scanners.
• Harmonization parameters derived from NEMA IEC Body Phantom data, applied to clinical datasets.
• Spearman's rank correlation coefficient (ρ) assessed rank preservation pre- and post-harmonization.

Main Results:

• Phantom-based harmonization reduced inter-scanner differences and enabled consistent clinical evaluation.
• SUVpeak demonstrated the highest rank preservation (ρ = 0.94-1.00) across tumor types and lesion sizes.
• Metabolic Tumor Volume (MTV) showed high rank correlation with specific thresholds but varied with definition, tumor type, and size; tumor-to-liver ratios had moderate preservation.

Conclusions:

• SUVpeak is the most robust metric for preserving lesion ranking after FDG PET/CT harmonization, suitable for multicenter studies.
• Careful standardization is crucial when utilizing MTV-based metrics for prognostic evaluation in multicenter settings.