RT Journal Article
SR Electronic
T1 Initial evaluation of a direct 4D PET parametric image reconstruction method
JF Journal of Nuclear Medicine
JO J Nucl Med
FD Society of Nuclear Medicine
SP 1462
OP 1462
VO 50
IS supplement 2
A1 Jianhua Yan
A1 Beata Planeta-Wilson
A1 Richard Carson
YR 2009
UL http://jnm.snmjournals.org/content/50/supplement_2/1462.abstract
AB 1462 Objectives The goal of this study is to evaluate an algorithm for direct reconstruction of PET parametric images from a one-tissue (1T) kinetic model. Methods Replicates of a 10-cm spherical phantom with 3 regions (Gray matter (GM), White matter(WM) and Basal Ganglia(BG)) were simulated; each comprised a 1-h list mode file (~4x108 events) based on a measured input function, typical kinetic parameters, and C-11 decay. Parametric images were produced with a frame-based method (FM) and the direct 4D method. The FM reconstructed individual frames (21 frames: 6x30s, 3x60s, 2x120s, 10x300s) and estimated each voxel's kinetic parameters from its time-activity curve with weighted least squares (weights based on noise equivalent counts). The 4D method, an extension of the MOLAR algorithm, was derived from a new likelihood function which incorporated the 1T model into the physical projection model, and employed a novel EM algorithm to estimate the kinetic parameters directly from the list-mode data. Both methods used ordered subsets (30 subsets, 2 iterations). Evaluation was based on each region’s bias and coefficient of variation (COV) across replicates. Results Percent bias of both methods were small, with < 6% for all parameters (K1, k2 and VT), although 4D was slightly worse than FM. 4D produce substantially lower variability than FM for all regions. After 2 iterations, K1 COV reduced from 5% to 3% for BG, 8% to 5% for WM, and from 7% to 4% for GM (~40% reduction); VT COV reduced from 5% to 3% for BG, from 7% to 4% for WM, and from 4% to 3% for GM. Conclusions This study indicates that the direct 4D reconstruction method is a promising approach to reduce noise in parametric images. Additional evaluation both by simulation and with real data is required.