Accuracy and Precision of Compartmental Model Parameters Obtained From Directly Estimated Dynamic SPECT Time-Activity Curves

B.W. Reutter, G.T. Gullberg, R.H. Huesman
2004 IEEE Transactions on Nuclear Science  
Quantitative kinetic analysis of dynamic cardiac single photon emission computed tomography (SPECT) data has the potential to provide better contrast between healthy and diseased tissue, compared to static images. However, imaging a rapidly changing radiopharmaceutical distribution with the use of a moving gantry yields inconsistent projection data that can generate artifacts in a time sequence of conventional image reconstructions. The artifacts can lead to biases in kinetic parameters
more » ... d from the image sequence. This source of bias can be eliminated by estimating B-spline models for time-activity curves directly from the projections. In this study, we perform Monte Carlo simulations to determine how the polynomial order and initial time sampling of the splines affect the accuracy and precision of compartmental model parameters obtained from directly estimated time-activity curves. The Mathematical Cardiac Torso (MCAT) phantom is used to simulate a realistic 15 min dynamic 99m Tc-teboroxime patient study in which 10 million total events are detected. For a large volume of normal myocardium (250 cc), the relative bias of the uptake and washout parameter sample means does not exceed 0.22% when using cubic or quadratic splines that provide rapid initial sampling. The coefficient of variation is about 1%. For small (8.4 cc) myocardial defects that exhibit reduced uptake and accelerated washout, the relative bias and coefficient of variation increase to maximum values of about 16% and 49%, respectively. These levels of accuracy and precision allow the defects to be discriminated from the normal myocardium. Index Terms-dynamic single photon emission computed tomography (SPECT), fully four-dimensional (4-D) reconstruction, kinetic parameter estimation
doi:10.1109/tns.2003.823014 fatcat:qiiogpprubemraa7z5mkiyh5fi