Towards tracer dose reduction in PET studies: Simulation of dose reduction by retrospective randomized undersampling of list-mode data

Objective: Optimization of tracer dose regimes in positron emission tomography (PET) imaging is a trade-o between diagnostic image quality and radiation exposure. The challenge lies in dening minimal tracer doses that still result in sucien t diagnostic image quality. In order to nd such minimal doses, it would be useful to simulate tracer dose reduction as this would enable to study the eects of tracer dose reduction on image quality in single patients without repeated injections of dierent amounts of tracer. The aim of our study was to introduce and validate a method for simulation of low-dose PET images enabling direct comparison of dierent tracer doses in single patients and under constant 18 inuencing factors. Methods: F-uoride PET data were acquired on a combined PET/magnetic resonance imaging (MRI) scanner. PET data were stored together with the temporal information of the occurrence of single events (list-mode format). A predened proportion of PET events were then randomly deleted resulting in undersampled PET data. These data sets were subsequently reconstructed resulting in simulated low-dose PET images (retrospective undersampling of list-mode data). This approach was validated in phantom experiments by visual inspection and by comparison of PET quality metrics contrast recovery coecien t (CRC), background-variability (BV) and signal-to-noise ratio (SNR) of measured and simulated PET images for dierent activity concentrations. In addition, reduced-dose PET 18 18 images of a clinical F-FDG PET dataset were simulated using the proposed approach. Results: F-PET image quality degraded with decreasing activity concentrations with comparable visual image characteristics in measured and in corresponding simulated PET images. This result was conrmed by quantication of image quality metrics. CRC, SNR and BV showed concordant behavior with decreasing activity concentrations for measured and for corresponding simulated PET images. Simulation of dose- 18 reduced datasets based on clinical F-FDG PET data demonstrated the clinical applicability of the proposed data. Conclusion: Simulation of PET tracer dose reduction is possible with retrospective undersampling of list-mode data. Resulting simulated low-dose images have equivalent characteristics with PET images actually measured at lower doses and can be used to derive optimal tracer dose regimes.