Direct 4D slice-wise whole-body parametric PET image reconstruction for continuous bed motion acquisitions

Abstract

Whole-body (WB) parametric PET imaging has recently demonstrated its clinical potential in enhancing quantification by employing robust dynamic analysis on a series of fast WB passes obtained either with step-and-shoot (SS) or continuous bed motion (CBM) acquisition mode. In particular, the CBM method has been associated with enhanced lesion detectability over SS in bed overlap regions for indirect Patlak influx rate constant (K i images. However, indirect Patlak is not very robust to the high noise levels often encountered in clinical WB dynamic PET imaging, thus limiting its clinical adoption. In addition, the CBM data at each transaxial slice are acquired within different time frames, while the SS data from all slices of a bed can be assigned a single average scan time. Thus, in the absence of list-mode data, accurate time data management may be challenging for dynamic CBM data analysis. In this study, we introduce a novel direct 4D WB parametric PET sinogram-based ordered subsets expectation maximization (OSEM) reconstruction algorithm specifically designed for the robust generalized Patlak analysis of dynamic CBM clinical PET data. The 4D algorithm employs the optimization transfer method to nest multiple faster Patlak subiterations within each of the slower global iteration steps and, thus, accelerate the convergence rate. Furthermore, an average scan time is assigned to each transaxial slice to construct an accurate set of slice-dependent nested Patlak model matrices and therefore permit a highly quantitative and efficient slice-wise 4D reconstruction of the CBM data. The clinical evaluation of the 4D CBM versus the 4D SS method was performed on a WB dynamic (12 passes) 18F-FDG PET/CT clinical study by switching the two modes between the passes. The quantitative analysis of a suspected high uptake foci located in a bed overlap region indicated an increase of 40-50% in K i target values and 35-45% in K i target-to-background scores for the CBM 4D Patlak method