Personalized 3D dosimetry of 90Y radioembolization: Monte Carlo simulation study using GATE

Abstract

This work aims to compare absorbed dose accuracy in tumour, normal liver, and lungs between the Partition Model (PM) and Monte Carlo (MC) simulation using MIRD phantom and patient tomography, considering patient-specific factors like pre-treatment parameters, source distribution, and tumour location. It outlines the procedure for acquiring patient images and treatment parameters from a medical institute and prepared by importing, segmenting, and registering tomographic images using the 3D Slicer software. Subsequently, the processed images are imported into GATE for MC simulation, with various source distributions applied for each patient. Absorbed dose estimates are also obtained using MC simulations with the MIRD phantom in Geant4, utilizing treatment parameters exclusively. These absorbed dose estimations are then compared with those obtained from patient measurements (PM). To visualize the isodose distributions resulting from the MC simulation, the SlicerRT extension of 3D Slicer is employed. The results revealed that, in most cases, higher absorbed doses when compared to PM calculations for MIRD phantom MC simulation. This discrepancy is attributed to the PM's assumption of localized energy deposition, neglecting crossfire effects. In MC simulation, when SPECT images were considered for source distribution, PM calculations yielded considerably higher absorbed doses for patients 1, 4, and 5, while patients 2 and 3 had lower values due to variations in source distribution intensity within the SPECT images. Employing region VOI for source distribution generally resulted in higher absorbed doses for the tumour in PM calculations, except for patient 3. Normalizing the prescribed tumour dose to 120 Gy highlighted differences in dose distribution, particularly for patients 2, 3, and 4. Isodose distribution analysis revealed variations in dose concentration between the simulations, with some regions experiencing higher doses in Geant4 simulations. This study emphasizes the significance of individualized dosimetry for 90Y radioembolization therapy through the utilization of MC simulation, the MIRD phantom, and patient tomography images. It presents a comparative analysis of this approach against PM calculations.