Characterizing Bladder Filling Physiology with Long-Axial-FOV Dynamic PET/CT

Abstract

Bladder filling during dynamic long-axial field-of-view (FOV) [ ¹⁸ F]Fluorodeoxyglucose ( ¹⁸ [F]FDG) positron emission tomography (PET) imaging leads to motion-related misalignment between frames. As a first step toward this misalignment correction, we track urinary bladder expansion and frame-byframe shape changes. Starting from the final PET frame, our algorithm applies intensity normalization, distribution-based correction, and region growing to estimate bladder volume over time. Compared to automated multi-organ segmentation frameworks, namely TotalSegmentator and MOOSE, the proposed PET-based method provides more accurate and physiologically plausible segmentations. Using our final-frame masks as reference, Dice-Sorensen coefficients were 0.40 } 0.37 for TotalSegmentator and 0.70 } 0.26 for MOOSE. In 21 successfully segmented patients, bladder volume increased approximately linearly over time (R ² = 0.94[0.88, 0.97]), while fractional anisotropy decreased (Spearman’s P= −0.86[−0.96,−0.61]), indicating a trend toward a more spherical bladder shape during filling. The bladder was typically no longer traceable at frame 40 (4-4.5 min post-injection).