Automated long axial field of view PET image processing and kinetic modelling with the TurBO toolbox

Abstract

<h3>Purpose</h3><p>Long axial field of view (LAFOV) PET imaging requires extensive automation due to the large number of target tissues. Therefore, we introduce an open-source analysis pipeline (TurBO, Turku total-BOdy) for automated preprocessing and kinetic modelling of LAFOV [¹⁵O]H₂O and [¹⁸F]FDG PET data. TurBO enables efficient, reproducible quantification of tissue perfusion and metabolism at regional- and voxel-levels through automated co-registration, motion correction, CT-based region of interest (ROI) segmentation, image-derived input function (IDIF) extraction, and region-specific kinetic modelling.</p><h3>Methods</h3><p>The pipeline was validated with Biograph Vision Quadra (Siemens Healthineers) LAFOV PET/CT data from 21 subjects scanned with [¹⁵O]H₂O and 16 subjects scanned with [¹⁸F]FDG. Six CT-segmented ROIs (cortical brain gray matter, left iliopsoas muscle, right kidney cortex and medulla, pancreas, spleen and liver) were used to assess different levels of tissue perfusion and glucose metabolism.</p><h3>Results</h3><p>Model fits showed high quality with consistent estimates at regional and voxel-levels (R² > 0.83 for [¹⁵O]H₂O, R² > 0.99 for [¹⁸F]FDG). Manual and automated IDIFs were in concordance (R² > 0.74 for [¹⁵O]H₂O, and R² > 0.78 for [¹⁸F]FDG) with minimal bias (< 4% and < 10%, respectively). Manual and CT-segmented ROIs showed strong agreement (R² > 0.82 for [¹⁵O]H₂O and R² > 0.83 for [¹⁸F]FDG). Motion correction had little impact on estimates (R² > 0.71 for [¹⁵O]H₂O and R² > 0.78 for [¹⁸F]FDG) compared with uncorrected data.</p><h3>Conclusion</h3><p>The TurBO pipeline provides fully automated and reliable quantification for LAFOV PET data. It substantially reduces manual workload and enables standardized, reproducible assessment of inter-organ perfusion and metabolism.</p>