A harmonisation framework for multi-centre preclinical PET imaging across a national research infrastructure

Quantitative preclinical positron emission tomography (PET) imaging plays an important role in translational research, yet variability across scanners, sites, and analysis workflows remains a major barrier to robust multi-centre studies. In this work, we developed and evaluated a harmonisation framework across a national network of preclinical PET facilities. Approach. Six identical NEMA NU~4-2008 image quality phantoms were manufactured and distributed to six sites, where they were imaged using local acquisition and reconstruction protocols across multiple scanner platforms. Main Results. Substantial variability in contrast recovery, uniformity, and spill-over ratios was observed under site-specific conditions. Centralised automated re-analysis produced only minor reductions in variability, indicating that inter-user effects were relatively small, compared to site-specific protocol variations. In contrast, optimisation of reconstruction parameters led to a marked reduction in inter-scanner variability (an average of 60\% reduction across larger rods), particularly for voxel-wise maximum contrast recovery coefficients (CRC_max^vox), which are more sensitive to noise and reconstruction settings than the standard axial averaging NEMA CRC metric. Following harmonisation, the variability of CRC_max^vox approached and was lower that of conventional axial-averaged NEMA analysis (CRC_max^avg), with reductions ranging 7--33\% across the 1--5 mm rods, demonstrating improved quantitative consistency even at the voxel level. Significance. These findings show that scanner and site-specific protocol variations are the dominant factors for harmonisation in preclinical PET. By systematically evaluating acquisition, image reconstruction and analysis related sources of variability across a national network of heterogeneous preclinical PET systems, this work provides a practical framework for harmonisation. The proposed framework is conceptually aligned with clinical harmonisation initiatives such as EARL and provides a practical basis for the future development of harmonised performance ranges across preclinical scanners. This enables more reliable multi-centre studies and supports the use of quantitative, voxel-wise imaging biomarkers in translational research.