Improved Lu-177 SPECT Statistical Efficiency Using Torch Recon® Multi-Energy Monte Carlo Reconstruction
Abstract
Purpose
Quantitative Lu-177 SPECT imaging requires balancing statistical precision with clinically feasible acquisition times. Conventional reconstruction relies on triple-energy-window (TEW) scatter correction and is therefore effectively limited to the 208 keV photopeak, as the 113 keV emission is dominated by higher-order scatter that is difficult to model analytically. As a result, a substantial fraction of detected photons are discarded, reducing signal-to-noise ratio (SNR) and increasing variability. Monte Carlo (MC)–based reconstruction explicitly models photon transport and cross-window coupling. Torch Recon® implements a full MC system model in both forward and adjoint operators, enabling simultaneous reconstruction of the 113 keV and 208 keV photopeaks. This study evaluates whether MC-based joint multi-energy reconstruction improves statistical efficiency in Lu-177 SPECT.
Methods
A NEMA IEC phantom (10–37 mm spheres; 0.89 MBq/mL; 9:1 contrast) was scanned 52 consecutive times. Reconstructions included (i) conventional analytical reconstruction (PyTomography, 208 keV), (ii) hybrid-MC reconstruction (Torch Recon® with analytical backprojection, 208 keV), and (iii) full-MC reconstruction (Torch Recon® using both 113 keV and 208 keV). Ordered-subsets expectation maximization (OSEM) was used in all cases. Precision was quantified as the percent standard deviation of sphere activity across acquisitions, and SNR was assessed using sphere-to-background contrast.
Results
Full-MC reconstruction substantially reduced background noise and improved SNR (e.g., 140.0 vs. 83.5 for full-MC and analytical reconstruction, respectively, in the 32 mm sphere) while maintaining comparable recovery coefficients. Hybrid-MC yielded modest improvements in precision, whereas full-MC produced pronounced improvements in both precision and SNR. These gains correspond to effective statistical speed-up factors consistent with scan-time reductions of approximately two-fold or greater.
Conclusion
Multi-energy Monte Carlo reconstruction implemented in Torch Recon® significantly improves precision and SNR in Lu-177 SPECT imaging. These improvements have the potential to enable meaningful scan-time reductions while preserving quantitative accuracy, supporting practical, patient-specific dosimetry workflows.