Whole-Body dynamic PET: Effect of temporal gaps on FDG Ki quantification from 3D and 4D reconstruction algorithms

Abstract

In whole-body dynamic PET imaging, acquisition is performed sequentially over bed positions to achieve whole-body coverage, introducing large temporal gaps in the acquired data. The objective of this work is to study the effect of temporal gaps on quantification of FDG net influx rate Ki from Patlak analysis using 3D and 4D reconstruction methods. A brain FDG scan was simulated for a single bed dynamic protocol and for 3 different cases of whole-body dynamic protocols. Reconstructions were performed using 3D and 4D MLEM algorithms. Two different dynamic models were considered for 4D reconstructions, a Patlak model for direct reconstruction of Ki and a spectral model for temporal regularisation. Except in the case of the direct Patlak 4D reconstruction, post reconstruction least squares fitting of the Patlak model was performed on the reconstructed frame activity images, at both the ROI and voxel level. Ki estimates derived from ROI level analysis showed similar level of bias for all simulations, and all reconstruction methods evaluated. In contrary estimates from voxel level analysis (parametric imaging) resulted to higher bias and noise for the whole-body protocols with 3D reconstruction, compared to the single bed protocol. For all whole-body protocols both 4D reconstruction methods outperformed 3D, by providing comparable bias and RMSE to the single bed case with 3D reconstruction. In all protocols with temporal gaps, evaluation of the standard-deviation bias trade-off showed that for 3D reconstruction standard deviation rises fast, with increasing iterations, with no convergence of the mean Ki estimate to a level of bias, while in contrary 4D reconstructions provided results with smaller standard deviation at matched bias and converged to a certain level of bias. These results suggest that behaviour of 4D reconstruction algorithms is preferable for use in FDG Ki quantification in parametric imaging, particularly for the case of whole-body dynamic imaging.