Comparison of Quantitative Monte Carlo Versus Osem Based SPECT Reconstructions for 177lu Using Two Energy Windows
Abstract
Purpose
Accurate 177Lu SPECT quantification is essential for patient-specific dosimetry in radiopharmaceutical therapies. Monte Carlo (MC) based reconstruction uses comprehensive system modeling to stochastically simulate photon transport and detector interactions, potentially improving quantitative accuracy and enabling more effective use of both the 113- and 208-keV photopeaks. However, evaluations of MC reconstruction for 177Lu remain limited.
Methods
The MC reconstruction algorithm was implemented using SIMIND integrated within the open-source PyTomography toolbox. OSEM reconstructs in counts and relies on an external calibration factor (CF) that can introduce bias, while the MC method directly estimates activity in MBq. Performance was evaluated using recovery coefficients (RCs). First, simulated 1-hour post-injection 177Lu-DOTATATE SPECT data from Siemens Symbia scanner were analyzed. For tumor quantification, OSEM CFs were derived from simulated point sources placed in attenuation phantoms of varying sizes for comparison. The MC method was then validated with an anthropomorphic phantom.
Results
The reconstructed activity map shows that our MC based reconstruction yields lower noise, especially at 113 keV, as confirmed by line profiles with fewer artificial fluctuations than OSEM. RC results at 113 keV vary substantially with the attenuation-phantom size used for OSEM calibration, largely due to increased impact of scatter. Although applying an ROI mask improves 208 keV accuracy, the 113 keV RC still show at least a 25% positive bias. In contrast, MC reconstruction provides stable and accurate RC (bias: 2.54% at 208 keV and 6.54% at 113 keV after 25 iterations). For the anthropomorphic phantom data, a central liver ROI was used to avoid partial-volume effects. Both photopeaks converge to similar RC, indicating good cross-energy consistency.
Conclusion
Compared with conventional OSEM, MC reconstruction offers substantial improvements in noise suppression and scatter correction, enabling reliable quantitative use of the highly scatter-affected 113 keV photopeak and potentially reducing the required scan time.