Automated Total-Body Perfusion Imaging with 15O-Water PET Using Basis Functions and Organ-Specific Model Selection.

Abstract

Long-axial-field-of-view PET with ¹⁵O-water allows perfusion to be measured in the whole body simultaneously. The purpose of this work was to describe a method for automated computation of total-body parametric perfusion images using PET information only and to validate the perfusion and volume of distribution (V _T) values obtained by comparing them with the gold standard (nonlinear regression analysis). Methods: Data from 10 subjects at Turku PET Centre were evaluated. Each subject underwent a 4-min 40-s dynamic PET/CT scan starting simultaneously with a controlled bolus administration of 350 MBq of ¹⁵O-water. Arterial and venous blood curves were defined using cluster analysis. Delay correction was performed by down-sampling the PET volume, using nonlinear regression for estimation of the delay for each subvolume, interpolation of delay values to the original matrix, and delay correction of all voxel time-activity curves, allowing for linearization of the model. Total-body perfusion images were calculated using several basis function implementations of the single-tissue-compartment model, considering the variations in blood supply to different organs. Model selection for each voxel was performed using cluster analysis to identify different organs. Perfusion and V _T values based on the automated parametric imaging method were validated by comparison of mean organ values with nonlinear regression of the appropriate compartment models to whole-organ time-activity curves. Results: The results showed good agreement between the parameters achieved from the automated parametric images and nonlinear regression. Correlation (R ²) and agreement between linear and nonlinear analyses were high, with an R ² of 0.99 for both perfusion and V _T, with a slope of 0.98 and 1.01 for perfusion and V _T, respectively. Conclusion: Perfusion and V _T values based on automated total-body parametric analysis agreed well with those based on nonlinear regression of whole-organ time-activity curves.