Abstract
503
Objectives: Early-dynamic 18F-FDG PET has the potential to derive blood flow without the need for a flow-specific radiotracer. All existing studies used a temporal resolution of 5-10 seconds and standard compartmental modeling that assumes time-constant kinetics. The FDG blood-to-tissue delivery rate K1 was considered as a surrogate of blood flow. However, this method can be only effective for regions such as aggressive tumors that have a high FDG extraction fraction and is ineffective for regions of low FDG-extraction. We propose to use high temporal-resolution (HTR) dynamic data acquisition to divide the tracer uptake period into a capillary transport phase and a tissue uptake phase. Blood flow can then be modeled independently of FDG extraction by using a time-varying kinetic model. The objective of this study is to test the validity of this model in HTR early-dynamic FDG-PET data. Methods: Twelve dynamic 18F-FDG PET/CT patient scans with the kidneys in the field of view of PET were included in this study. Dynamic PET data were acquired in list-mode format and binned into HTR time frames (2 seconds per frame) for the early 2 minutes post tracer injection. Regional time activity curves were extracted from dynamic PET images in the region of renal cortex and fit using two different kinetic models: standard one-tissue compartmental model with time-constant kinetics and a new time-varying compartmental model that allows different kinetic constants in the vascular phase and tissue phase. The transit time Tc, blood flow F and extraction fraction E were estimated from fitting of the time activity curve using the derived analytical model equation and a basis-function optimization algorithm. FDG delivery rate K1 was calculated by K1=FE. These two kinetic models were compared for curve fitting using the Akaike information criterion (AIC) and F test. Statistical association of different kinetic parameters was analyzed using the Pearson correlation analysis. Results: The time-varying kinetic model fit the HTR time activity curves more closely than standard time-constant kinetic model, particularly for the peak values. Statistical quality metrics showed the time-varying model achieved lower AIC and higher F values in all patients, meaning better fit quality by the new model. Renal extraction fraction of FDG was relatively low (E=35±10%). The estimated transit time was Tc=11±3 seconds. Renal FDG K1 was 0.65±0.14 mL/g/min and blood flow F was 1.94±0.53 mL/g/min by the time-varying model. FDG K1 by the time-varying model was correlated with K1 by the traditional time-constant compartmental model (r=0.68, p=0.02). Neither of the two K1 estimates correlated with blood flow F (r=0.23, p=0.48; r=0.20, p=0.54), indicating that FDG K1 and blood flow reflect different physiological processes in renal cortex. Conclusions: HTR early-dynamic FDG PET data favors time-varying compartmental modeling over standard time-constant compartmental modeling. This new method allows separate modeling of blood flow and FDG K1 and offers the promise of adding quantitative perfusion imaging to existing metabolic FDG-PET imaging to enable single-tracer multiparametric imaging. This study focused on renal cortex for demonstrating the feasibility but the method is applicable to other organs and in principle to any intravenously delivered radiotracer. While extension of the method for parametric imaging can be challenging because of high noise associated with individual image voxels, the emerging EXPLORER scanner has a much higher sensitivity than standard clinical PET scanners and provides a great opportunity for voxel-wise implementation of time-varying kinetic modeling. Our future work will exploit this modeling approach for total-body perfusion imaging on the EXPLORER and validate it against standard blood flow tracers. Research Support: This work is supported in part by NIH R21 HL 131385.
