Abstract:521 Objectives: Glucose transport is a vital component of the metabolic process. The rate of glucose transport from plasma to tissue cells can be quantified using dynamic 18F-FDG PET via the estimation of the K1 parameter in compartmental modeling. However, the accuracy of K1 quantification may be compromised by the assumption of standard compartmental modeling that radiotracer is instantaneously mixed in the blood compartment and ignores that a capillary flow period may exist before the radiotracer is actually transported into tissue cells. In this study, we use high-temporal resolution (HTR) dynamic imaging with a newly developed time-varying kinetic model to better quantify glucose transport. Methods: Two FDG-PET studies were conducted. Prior IRB approval and informed consent were obtained in all cases. The first study involved 14 patient scans with the kidneys in the field of view of a standard PET scanner (GE Discovery 690). Dynamic PET data were reconstructed into HTR 2-s time frames for the first 2 minutes post injection. Regional time activity curves (TACs) were extracted from dynamic images in the renal cortex. The TAC from the descending aorta was used as the image-derived input function (IDIF) for kinetic modeling. The second study was performed using the EXPLORER, a 2-meter long high-sensitivity PET scanner, capable of total-body dynamic imaging. A healthy human subject was scanned and the data of the first 3 minutes were reconstructed into 110 HTR time frames (60x1s, 30x2s, 20x3s). Regional TACs were extracted from regions of interest (ROI) including renal cortex, white matter, gray matter, and myocardium. The left ventricle TAC was used as the IDIF. In both studies, regional TACs were fitted using the standard one-tissue compartmental model with time-invariant kinetics and a new time-varying model which explicitly incorporates the blood flow component in addition to the glucose transport rate K1. These two models were compared for TAC fitting using the Akaike information criterion (AIC). Results: In the first study, the new time-varying model fit the renal TACs more closely than the standard time-invariant model, particularly for the peak values. The corresponding impulse response functions of the two fits indicated the difference and connection between conventional K1 by the standard model and flow-corrected K1 by the new model. Conventional K1 values were 1.09±0.18 ml/g/min in renal cortex. In comparison, the flow-corrected K1 estimates were 0.55±0.15 ml/g/min. The reduction is due to the fact that the new model takes into account the effect of renal blood flow (estimated values: 1.54±0.40 ml/g/min) in the kinetic parameter estimation. The result also indicates conventional FDG K1 estimation is likely a mix of glucose transport rate and blood flow. The time-varying model achieved lower AIC values in all patients, indicating better fit quality by the new model. In the second study using EXPLORER, the time-varying model also better fitted the TACs than the standard model according to both visual fit quality and statistical AIC evaluation. Compared with the flow-corrected K1 estimates, conventional K1 estimates were 40%-60% higher in the renal cortex and myocardium and about 5% higher in gray matter and white matter. The AIC difference further confirmed the new model provided a better fit for different organ ROIs. Conclusion: Glucose transport rate K1 may be overestimated by conventional compartmental modeling due to the effect of partial blood flow involved in the kinetic parameter estimation. With HTR dynamic imaging, this overestimation of K1 can be corrected using a time-varying model. While there was no ground truth available for validation in our studies, the statistical fit evaluation using AIC suggests the new model is more appropriate for quantification of glucose transport. Our future work will further validate the method by correlating flow-corrected K1 with the expression of glucose transporters in cancer and fatty liver disease.

Blind identification of the kinetic parameters in three-compartment models

Robust Reconstruction Of Physiological Parameters From Dynamic Pet Data

Exact parameter identification in PET pharmacokinetic modeling using the irreversible two tissue compartment model

Noninvasive Estimation of Macro-parameters for Dynamic PET Images Using Deep Learning

Pharmacokinetic analysis of tissue microcirculation using nested models: multimodel inference and parameter identifiability

A physiology--based parametric imaging method for FDG--PET data

Robust estimation of kinetic parameters in dynamic PET imaging

Compartmental analysis of dynamic nuclear medicine data: models and identifiability

Simultaneous Estimation and Segmentation from Projection Data in Dynamic PET.

Non-invasive kinetic modelling approaches for quantitative analysis of brain PET studies

Multiparametric Cardiac 18F-FDG PET in Humans: Kinetic Model Selection and Identifiability Analysis

Non-invasive kinetic modelling of PET tracers with radiometabolites using a constrained simultaneous estimation method: evaluation with 11C-SB201745

State-Space Reconstruction of Pet Parametric Maps

Model Corrected Blood Input Function to Compute Cerebral FDG Uptake Rates From Dynamic Total-Body PET Images of Rats in vivo

Novel Method to Estimate Kinetic Microparameters from Dynamic Whole-Body Imaging in Regular-Axial Field-of-View PET Scanners

Patient-adaptive lesion metabolism analysis by dynamic PET images.

Quantification of Glucose Transport Using High Temporal Resolution Dynamic PET Imaging

Identifiability of spatiotemporal tissue perfusion models

Reinforcement Learning Based Model Selection and Parameter Estimation for Pharmacokinetic Analysis in Drug Selection.

Noninvasive Estimation of Macro-Parameters by Deep Learning

Nonparametric Residue Analysis of Dynamic PET Data With Application to Cerebral FDG Studies in Normals