Abstract
206
Introduction: The world’s first 2-meter long PET/CT scanner (uEXPLORER) has been developed to offer ultra-sensitive total-body molecular imaging for biomedical research and clinical applications. It provides opportunities for tracer kinetic analysis of the human body in terms of physiology, biochemistry and pharmacology. We have previously developed a linear parametric imaging method using the Patlak model. In this work, we investigate a nonlinear kinetic analysis method based on a two-tissue compartment model and compare the results with the Patlak model for total-body parametric imaging using dynamic PET data. Methods: We conducted human total-body dynamic PET study using the uEXPLORER scanner. A one-hour 18F-FDG scan was performed and the dynamic dataset was divided into temporal frames to capture the change of tracer distribution over time with protocol of 60×1 s, 30×2 s, 20×3 s, 12×10 s, 50×30 s, and 15×120 s. Quantitative image reconstruction was performed using our total-body kernel-based algorithm with corrections for attenuation, detector response, scatters and randoms. We implemented total-body parametric imaging using the nonlinear two-tissue compartment method (2TC). The irreversible model was adapted on FDG kinetics for glucose metabolism. and for comparison with the linear Patlak model. The dynamic data of the last 30-min of the scan and the full 60-min of dynamic data were used for Patlak and 2TC analyses, respectively. An image derived input function was extracted using the time activity curves (TAC) from a region placed over the descending aorta. Parametric images of the FDG influx rate Ki and distribution volume b were generated using both Patlak and 2TC models and compared. Regions of interest (ROIs) were drawn in the gray matter (GM), white matter (WM), myocardium and liver. The mean and standard deviation of the estimated parameters over voxels within each ROI were calculated and compared. The correlation between the Patlak and 2TC results was also investigated.
Results: Total-body parametric images reconstructed by the Patlak and 2TC models showed good image quality. Both models fitted the last 30-min of the TACs well. The Ki values estimated by 2TC and Patlak models were highly correlated (R2=0.98), while also demonstrated heterogeneity across regions. The correlation for the intercept b was lower because of the intrinsic model difference. ROI analysis showed that the mean values of the estimated Ki between the two models were close. The 2TC model resulted in lower spatial variation within each ROI because it used the full dynamic scan. However, the 2TC model was also sensitive to subject motion during the scan and resulted in outliers near the boundaries of brain, arms, and legs. In contrast, the Patlak estimation was insensitive to subject motion during the first 30-min of the scan and resulted in far fewer outliers. Conclusion: We have demonstrated total-body parametric imaging using nonlinear kinetic analysis and cross-validated with the Patlak model. Our results show that the 2TC model produces parametric images with lower noise than the Patlak model, but is more susceptible to artifacts caused by subject motion during the scan. Acknowledgements: Support for this work includes NIH grant R01 CA206187 and a UC Davis Innovative Development Award. We acknowledge the contributions of all team members from UC Davis, United Imaging Healthcare and Zhongshan Hospital.
