Iterative Reconstruction with Monte Carlo Based System Matrix for Dedicated Breast PET

Increasing sensitivity by reducing the ring diameter of breast PET system may degrade image performance at field of view periphery. In this work, the authors present a framework for computing and incorporating an accurate system model of breast PET utilizing GATE Monte Carlo simulation to compensate for this performance degradation. The system matrix (SM) generation count statistics was maximized by taking into account the geometric symmetry of the scanner. The SM was incorporated into MLEM reconstruction and compared with the Siddon ray-tracing algorithm to evaluate point source resolution and contrast recovery coefficient (CRC) of hot spheres at various radial locations. Both spatial resolution and CRC using SM based MLEM was approximately position invariant, whereas the CRC and spatial resolution with Siddon based MLEM was substantially lower for locations near the periphery of the FOV. The CRC vs noise tradeoff was markedly better with the SM based MLEM.