Calibration-Free Quantitative SPECT Imaging of Alpha Emitters Using Torch Recon®: Monte Carlo-Based Reconstruction Model
Abstract
Purpose
Quantitative SPECT imaging of the alpha-emitters Ac-225 and Pb-212 is fundamentally limited by a low count rate and high-energy photon emissions that produce collimator penetration, septal scatter, and detector backscatter within clinically used photopeak windows. Conventional triple-energy-window (TEW)–based scatter correction does not adequately model these effects, leading to biased activity estimates. We present a fully Monte Carlo (MC)–based reconstruction framework that explicitly models these physical processes and inherently accounts for system sensitivity, enabling direct reconstruction in activity units without calibration scans, which can reduce study duration, occupational radiation exposure, and radionuclide preparation requirements in clinical studies.
Methods
Torch Recon® incorporates a physics-based MC SPECT system model into the forward and adjoint operators of an ordered-subsets expectation–maximization (OSEM) algorithm, enabling joint multi–energy-window reconstruction. The MC-based adjoint explicitly models cross-window coupling from photon penetration and scatter. Model accuracy was validated using point-source measurements and independent SIMIND simulations, while reconstruction performance was evaluated using NEMA phantoms for Ac-225 and Pb-212. For Ac-225, 218 keV and 440 keV windows were jointly reconstructed to estimate separately the distributions for the Ac-225→At-217 and Bi-213→Bi-209 components of the decay chain. For Pb-212, MC-based reconstructions were compared with conventional TEW-based reconstruction implemented in PyTomography.
Results
For Ac-225, joint reconstruction of the decay components produced stable activity estimates, with expanded ROI recovery near 100% for larger spheres in both cases. For Pb-212, MC-based reconstruction demonstrated superior hot-sphere recovery and background suppression relative to conventional reconstruction. Whereas the conventional approach required external calibration and overestimated activity in expanded ROIs, the MC-based method achieved total recovery coefficients near unity and substantially higher sphere-to-background contrast.
Conclusion
Fully MC-based forward and adjoint modeling enables accurate, calibration-free quantitative SPECT imaging of alpha emitters, improving robustness to scatter and penetration effects and supporting reliable patient-specific dosimetry for targeted alpha therapy.