TAME-Q: An Open-Source Preprocessing Pipeline for Reproducible Semi-Quantification of Florzolotau (18F) PET

Purpose Accurate semi-quantification of florzolotau (18F) PET in Alzheimer’s disease (AD) and non-AD tauopathies is essential. While histogram-based approaches are promising, their implementation relies on proprietary software, limiting accessibility. Thus, we developed TAME-Q (Tomographic Image Preprocessing Using Automated Multi-Reference Estimation for Quantification), an open-source preprocessing pipeline that performs histogram-based semi-quantification. This study aimed to use this open-source framework to produce results consistent with those of prior work and to validate the pipeline’s performance for florzolotau PET semi-quantification. Methods This study included 37 individuals on the AD continuum, 46 with probable progressive supranuclear palsy-Richardson syndrome (PSP-RS), and 50 healthy controls. TAME-Q was used to co-register T1-weighted MR and florzolotau PET images and perform semi-quantification by fitting a bimodal Gaussian mixture to grey-matter signal intensities. Elastic Net models yielded AD-tau and PSP-tau scores indicating disease-characteristic tau burdens. The standardized uptake value ratios (SUVRs) of the inferior temporal gyrus (ITG) and globus pallidus (GP) were compared between groups. The two pipelines were compared via correlation analysis and ROC/AUC comparison using DeLong’s test. Results TAME-Q processed all subject data without manual correction. The SUVR was elevated specifically in the ITG for AD and the GP for PSP. Tau scores correlated with previous scores (AD-tau: R = 0.983; PSP-tau: R = 0.966) and had AUCs of 0.999 (AD) and 0.977 (PSP-RS) with no significant difference in discriminative performance. Conclusion TAME-Q enables fully automated and reproducible semi-quantification of florzolotau PET comparable to previous methods. Its open-source implementation supports transparent dissemination and broader adoption in research settings.