Region resolvability versus noise level characteristics for joint spatial and kinetic parameter estimation in inconsistent projection dynamic ECT

For emission computed tomography (ECT) studies of temporally static source distributions, well-known guidelines exist for the number of resolution elements which may be acceptably resolved in a reconstructed image, at a given noise level. Owing to the incomplete angular-temporal sampling in such ECT modalities as rotating camera dynamic single photon ECT (SPECT), the acquired sinogram is not a consistent representation of a Radon transform, and consequently, no analogous bounds on the performance of dynamic reconstruction algorithms may be derived from Radon transform theory. Applying what the authors believe to be the first spatiokinetic parameter estimation algorithm able to simultaneously estimate both the geometry and kinetics of multiple dynamic regions directly from inconsistent projections, the authors establish empirical estimates for the number of regions whose boundaries and time-activity curves (TAC's) may be simultaneously estimated to a specified degree of accuracy at a given signal-to-noise ratio (SNR), Surprisingly, we find that regional TAC recovery for a segmented annulus myocardial phantom is relatively insensitive to noise at realistic SNR's and to a twofold increase in the number of resolution elements. It is concluded that errors in the recovered regional TAC's are due primarily to the poorly conditioned nature of the spatiokinetic parameter estimation problem.