A unified approach to 4D gated myocardial SPECT reconstruction

The authors study similarities between proposed methods for noise control in 4D gated myocardial perfusion SPECT reconstruction. Three methods have been proposed: Fourier filtering in four dimensions, space-time Gibbs priors in 4D MAP reconstruction, and Karhunen-Loeve (KL) filtering along the time axis with 3D MAP reconstruction. The authors demonstrate that all 3 are cases of 4D MAP reconstruction with space-time priors. Through simulation experiments with a realistic beating heart phantom, they show that nonquadratic space-time Gibbs priors produce less error in edge location and volume estimation than the other two methods for the same degree of noise smoothing. The authors attribute this to the time-global nature of the Fourier and KL smoothing versus the time-local nature of the space-time Gibbs model. They conclude that the 4D MAP method is most effective in their study, but that there are many variables deserving of further study