Fully 4D reconstruction of dynamic SPECT images based on the estimation of spatiotemporal basis coefficients directly from projection measurements

Abstract

Quantitative kinetic analysis of dynamic cardiac SPECT data provides unique information that can enable improved discrimination between healthy and diseased tissue, compared to static imaging. Conventional kinetic parameter estimation is based on the reconstruction of a dynamic sequence of 3D images from complete projections and the fitting of the compartmental kinetic models to time-activity curves generated by overlaying regions of interest on the reconstructed sequence. The problem with this approach is that, due to the gantry motion involved and the changes in the distribution of radiopharmaceutical during the acquisition, the projection data at different angles come from ditTerent tracer distributions resulting in inconsistent projections. Images reconstructed from these inconsistent projections can contain artifacts that lead to bias in the estimated kinetic parameters. In this work we present a fully 4D reconstruction algorithm for dynamic SPECT in a parallel ray geometry. The method is based on modeling the spatial and temporal distribution of the radiotracer inside the field of view with spatial tri-linear B-spline and 4 piecewise quadratic temporal B-spline basis functions The algorithm was tested on a dynamic 99mTc-teboroxime cardiac SPECT/CT canine study.