TY - JOUR T1 - Comparison of spectral pharmacokinetic image reconstruction algorithms JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 1461 LP - 1461 VL - 50 IS - supplement 2 AU - Andrew McLennan AU - Michael Brady Y1 - 2009/05/01 UR - http://jnm.snmjournals.org/content/50/supplement_2/1461.abstract N2 - 1461 Objectives Parametric images such as volume of distribution and rate of uptake are powerful ways to portray spatio-temporal tracer distributions in vivo. Conventionally, each contiguously acquired time frame is reconstructed independently (with or without spatio-temporal smoothing) before kinetic modelling is applied as a post-reconstruction step. Our aim here is to compare various direct spectral kinetic parameter estimation algorithms applied directly to the sinogram data. Methods We reformulate the standard independent timeframe OSEM and ART iterative reconstruction algorithms such that the parameters being estimated are the weights of exponential bases which when convolved with a known input function form the dynamic tracer time activity curves. As the number of non-zero coefficients for each voxel would ideally correspond to its particular model order (i.e. number of tissue compartments), various threshold-based sparsity constraints are compared. Results Explicitly incorporating consistency between neighbouring temporal frames removes the need to first form a dynamic sequence of noisy images. Macro parameters are computed for each algorithm and threshold method to determine the effect the sparsity constraint has on the parameter accuracy. Boundaries between regions of interest are sharper and projected dPET images have significantly higher SNR and lower MSE, with more accurate macro parameters. The sparsity constraints have significant influence on the macro parameters and need to be selected carefully. Conclusions Direct calculation of pharmacokinetic parameters from sinogram data enables more accurate image volumes estimates and lower SNR parameters than when estimated from pre-reconstructed images. ER -