Introducing time-of-flight and resolution recovery image reconstruction to clinical whole-body PET parametric imaging

We recently proposed a dynamic multi-bed acquisition scheme enabling whole-body FDG PET parametric imaging from limited axial field-of-view PET/CT scanners in clinically feasible scan times. However, the proposed framework was only evaluated for standard ordered subsets expectation maximization (OSEM) reconstruction. Currently, state-of-the-art commercial PET/CT scanners are equipped with advanced detection systems, capable of measuring the time-of-flight (TOF) of each annihilated photon enabling to confine the location of the annihilation position to a small segment within the line of response. As such, noise propagation is reduced and TOF reconstruction may provide superior contrast to noise ratio (CNR). Furthermore, image reconstruction is enriched with the feature of scanner resolution point spread function (PSF) modeling within the system response matrix of OSEM algorithm, similarly allowing for higher CNR. In this study, we extended TOF and PSF modeling to the dynamic multi-bed domain and systematically investigated their impact on the quality of wholebody PET parametric images. The state-of-the-art Siemens Biograph mCT scanner and its reconstruction suite were utilized. An extensive set of realistic 4D phantom simulations for the mCT scanner with and without TOF features were performed. Resolution degradation was applied to match a spatial resolution of 4.5mm. Then, TOF and non-TOF reconstructed images with and without resolution modeling were produced. Subsequently, the impact of TOF and PSF was assessed for standard and generalized Patlak models. Our results demonstrate the potential benefit of introducing TOF and PSF in parametric imaging, with both features providing superior noise vs. bias trade-off. Tumor-to-background ratio is enhanced by 30% when utilizing TOF, while CNR is improved by 40% and 60% when either TOF or PSF capabilities are introduced, respectively. Finally, total CNR enhancement approaches 100% if the two features are combined.