Uncertainty Propagation from Projections to Region Counts in Tomographic Imaging: Application to Radiopharmaceutical Dosimetry

Lucas Polson,Sara Kurkowska,Chenguang Li,Pedro Esquinas,Peyman Sheikhzadeh,Mehrshad Abbasi,Francois Benard,Carlos Uribe,Arman Rahmim

2024-10-05

Abstract:Radiopharmaceutical therapies (RPTs) present a major opportunity to improve cancer therapy. Although many current RPTs use the same injected activity for all patients, there is interest in using absorbed dose measurements to enable personalized prescriptions. Image-based absorbed dose calculations incur uncertainties from calibration factors, partial volume effects and segmentation methods. While previously published dose estimation protocols incorporate these uncertainties, they do not account for uncertainty originating from the reconstruction process itself with the propagation of Poisson noise from projection data. This effect should be accounted for to adequately estimate the total uncertainty in absorbed dose estimates. This paper proposes a computationally practical algorithm that propagates uncertainty from projection data through clinical reconstruction algorithms to obtain uncertainties on the total measured counts within volumes of interest (VOIs). The algorithm is first validated on ${}^{177}$Lu and ${}^{225}$Ac phantom data by comparing estimated uncertainties from individual SPECT acquisitions to empirical estimates obtained from multiple acquisitions. It is then applied to (i) Monte Carlo and (ii) multi-time point ${}^{177}$Lu-DOTATATE and ${}^{225}$Ac-PSMA-617 patient data for time integrated activity (TIA) uncertainty estimation. The outcomes of this work are two-fold: (i) the proposed uncertainty algorithm is validated, and (ii) a blueprint is established for how the algorithm can be inform dosimetry protocols via TIA uncertainty estimation. The proposed algorithm is made publicly available in the open-source image reconstruction library PyTomography.

Medical Physics