Attenuation Correction in SPECT during Image Reconstruction using an Inverse Monte Carlo Method: A Simulation Study

Introduction: The main goal of SPECT imaging is to determine activity distribution inside the organs of the body. However, due to photon attenuation, it is almost impossible to do a quantitative study. In this paper, we suggest a mathematical relationship between activity distribution and its corresponding projections using a transfer matrix. Monte Carlo simulation was used to find a precise transfer matrix including the effects of photon attenuation.  Material and Methods: List mode output of the SIMIND Monte Carlo simulator was used to find the relationship between activity distribution and pixel values in projections. The MLEM iterative reconstruction method was then used to reconstruct the activity distribution from the projections. Attenuation-free projections were also simulated. Reconstructed images from these projections were used as reference images. Our suggested attenuation correction method was evaluated using three different phantom configurations: uniform activity and uniform attenuation phantom, non-uniform activity and non-uniform attenuation phantom, and NCAT torso phantom. The mean pixel values and fits between profiles were used as quantitative parameters. Results: Images free from attenuation-related artifacts were reconstructed by our suggested method. A significant increase in pixel values was found after attenuation correction. Better fits between profiles of the corrected and reference images were also found for all phantom configurations.  Discussion and Conclusion: Using a Monte Carlo method, it is possible to find the most precise relationship between activity distribution and its projections. Therefore, it is possible to create mathematical projections that include the effects of attenuation. This helps to have a more realistic comparison between mathematical and real projections, which is a necessary step for image reconstruction using MLEM. This results in images with much better quantitative accuracy at a cost of computation time and memory.