PET image reconstruction and dosimetry from voxelized phantoms with GATE

Abstract

Accurate patient-specific internal dosimetry is a critical concern in the field of nuclear medicine. GATE is a robust Monte Carlo toolkit renowned for its integration of Geant4 algorithms, PET specialized tools and patient-specific dosimetry estimation. In this work, a GATE model is developed to simulate the PET scanner Biograph Vision and a voxelized phantom from Computed Tomography (CT) images. The segmentation of the CT images is performed using a deep learning model capable of automatically delineating anatomical structures, setting the basis for creating the patient-specific voxel phantom. GATE Nested parameterization method is employed for its efficient memory usage in defining geometry and faster navigation for ultra-large number of voxels. Simultaneously, PET acquisition data is used to assign the corresponding activity of a source to each voxel. This study aims to highlight the potential of GATE as a simulation tool within a methodology that integrates PET image reconstruction and internal dosimetry calculation, focused specifically on its application in prostate diagnostic testing via 18F-FDG. S-Value and dose are calculated for the prostate gland, yielding values of 1.52 E−4 mGy/MBq‧s and 8.1 mGy, respectively, consistent with literature findings. Differences in S-Values with the ICRP Phantom and with OpenDose for surrounding organs range from 0.5% to 67.9%, which can be attributed to the choice of phantoms used in calculations. This work confirms the capability of GATE to reproduce clinical studies using anthropomorphic voxelized models.