A Model-based Harmonization Strategy of Standard Uptake Values for Paired PET Studies

As both a diagnostic and quantitative monitoring tool, positron emission tomography (PET) imaging provides precise assessment for therapeutic efficacy in cancer management, neurological disorders, and cardiovascular diseases. In multicenter studies, the need for standardization of quantitative parameters of PET across systems is urgent. Systematic deviations from different scanners can compromise the consistency of imaging characteristics. The complex noise characteristics of clinical PET images, influenced by factors such as reconstruction protocols, scanner performance, and human factors, pose a challenge to the reliability of multicenter comparison studies. This study proposes a harmonization strategy aimed at improving the comparability of PET images from different sources. We present a model-based technique for normalizing PET images and assessing the measurement uncertainties in clinically relevant variables, specifically, SUV[Formula: see text]. Our approach successfully harmonizes SUV[Formula: see text] across different imaging modalities. We collected paired imaging data from 37 patients, obtained using both PET/CT and PET/MR scanners. The images show notable differences in statistical characteristics between scanners. The harmonization results show that the difference in SUV[Formula: see text] for phantom data decreased from the original value of 5.54± 5.43 to -0.06 ± 0.87, and the difference in SUV[Formula: see text] for patient data decreased from the original value of 3.95± 9.55 to -0.11± 1.10. Our method is able to improve the comparability of corresponding cases. It has demonstrated the efficiency of harmonization on paired PET images and suggests a potential application in multicenter PET studies.

The online version contains supplementary material available at https://doi.org/10.1007/s43657-025-00289-8.