Applications of Monte Carlo simulation in internal dosimetry of radiopharmaceuticals employed in nuclear medicine

Internal dosimetry (ID) is acquiring a fundamental role in nuclear medicine for the planning, optimization and monitoring of radiopharmaceutical therapies and diagnostics. Direct Monte Carlo (MC) simulation of radiation transport, using morphological and functional tomographic imaging as input data to model, respectively, patient’s body and radiopharmaceutical biodistribution, is the gold standard approach for ID. Despite being the most accurate and patient-specific method available, MC ID is not routinely employed in clinics, because of its relevant requirements in terms of computational resources and times, and thus other simplified approaches - such as MIRD formalism - are usually preferred instead. Anyhow, it is MC which enables to validate the applicability of simplified methods within specific conditions, and is the most reliable method for research in the field of ID. The research studies presented in this dissertation had the common denominator of taking advantage of voxel-level patient-specific MC simulation to obtain new results in ID for some nuclear medicine therapies and diagnostic exams. In parallel, focus was put in investigating the limits of such MC methodology, in developing viable corrections to overcome them, and in trying to optimize the computational times of these MC calculations, carried out with the GEANT4-based toolkits GATE and GAMOS. A first topic concerned MC ID of 18F-choline PET diagnostics cases and 90Y-microspheres TARE therapy cases, focused on the effect on dosimetric outcomes of reconstruction noise, background noise and motion blurring affecting the functional scans used as input data for the simulations. From these studies emerged the non-negligible influence of the mentioned noise effects, highlighted by implementing threshold-based and segmentation-based filtering techniques of the functional scans, which appeared an effective tool for the correction of absorbed dose artefacts caused by noise. A second topic regarding MC 90Y TARE dosimetry was addressed in a study aimed at reducing computational times while maintaining high dosimetric accuracy, which was carried out by investigating the effect of different simulation parameters, as the production cuts on secondary particles and the resolution of CT scans, and finding the best combinations of them. In addition, the OpenDose Dosimetry 3D module of the software 3D Slicer, developed by the OpenDose collaboration for the user-friendly implementation of voxel ID workflows, was tested and validated for 90Y TARE cases, comparing the results of the multiple algorithms offered, including MC, with literature results. Finally, an original simplified renal dosimetry method for 177Lu PRRT therapies, based on a single SPECT/CT scan and multiple external detector measurements, was designed and tested with the help of MC simulations. A proof-of-concept study for this protocol was indeed developed performing phantom experiments, reproducing them with MC simulations, and testing the proposed method on patient data. In the last step, simulating external measurements performed with collimated probe that would be used to deduce information on the radiopharmecautical biokinetics in kidneys, an estimate of the renal absorbed dose was retrieved adopting the simplified protocol and was compared to the estimate obtained with a full imaging-based MC voxel dosimetry, showing promising agreement. The proposed protocol would reduce the needed tomographic imaging to a single scan, minimizing machine occupation time and improving patient comfort.