Monte Carlo optimization of 4D smoothing methods for gated myocardial perfusion SPECT

We develop a technique for optimizing space-time smoothing methods proposed for gated myocardial perfusion SPECT. Gated SPECT synchronizes image acquisition with an ECG signal so that a time loop of 3D SPECT images is acquired. The optimization technique employs a realistic spline-based beating heart phantom. We simulate noise-free SPECT data from the phantom including all principal physical degrading effects, and then we simulate an ensemble of noisy datasets. Each member of the ensemble is processed using the candidate reconstruction method, varying the smoothing parameters to optimize bias and variance of a quantitative measure representing the task to be performed with the study. The method is demonstrated on the optimization of smoothing parameters for left ventricular volume estimation. We conclude that Monte Carlo simulation using realistic models of motion is essential in predicting and optimizing performance of 4D reconstruction methods.