Application of a spatially variant system model for 3-D whole-body pet image reconstruction

Accurate system modeling in tomographic image reconstruction has been shown to reduce the spatial variance of resolution and improve quantitative accuracy. System modeling can be improved through analytic calculations, Monte Carlo simulations, and physical measurements. This work presents a novel measured system model and incorporates this model into a fully 3-D statistical reconstruction method. Empirical testing of the resolution versus noise benefits reveal a modest improvement in spatial resolution at matched image noise levels. Convergence analysis demonstrate improved resolution and contrast versus noise properties can be achieved with the proposed method with similar computation time as the conventional approach. Images reconstructed with the proposed model contain correlated noise structures which are difficult to characterize with accepted NEMA noise metrics.