Implementation and automatization of validation tests for the simulation of Nuclear Medicine studies

Abstract

Nowadays, Nuclear Medicine techniques are of great importance in the clinical practice, especially in Neurology and Oncology, for both Diagnostics and Treatment. Focusing on imaging techniques, quantitative methods have shown to be an alternative to the traditional interpreting of images to give support to the clinician for a more accurate diagnostic. The lack of ground-truths has difficulted the implementation of these quantitative methods in the clinical practice, and consequently the use of simulation has raised as a solution for obtaining reference images. In particular, SimPET is a platform that aims to simulate and reconstruct Positron Emission Tomograph (PET) images while being able to model any scanner by introducing its parameters. The introduction of a new scanner requires a validation process to ensure it works in accordance of the manufacturer’s specifications. To do so, the NEMA protocols are a set of tests to allow a comparison with these specifications. In this work we are developing an implementation in Python and automatization of these tests to allow the easy validation of new scanners for SimPET. Four of the five specified tests were implemented, concerning the Spatial Resolution and the Sensitivity of the tomograph, and the Image Quality and the fraction of Scattering of the reconstructed images. The results were tested and compared with the experimental ones of the Discovery ST scanner. The results show that the values of Resolution and Sensitivity do not strictly coincide with the experimental ones, and further work on tuning the adjustments of the simulation need to be done. For the Image Quality test we obtained parameters that seem to be in accordance with the theoretical behaviour, but the lack of comparative results does not allow us to extract relevant conclusions. In the case of the Scatter Fraction test, we have shown that the suggested method does not work for very low activities. Thus, due to time and computational limitations further work needs to be done in the development of the techniques, but this work has shown the feasibility of its implementation.