Direct PET reconstruction of regional binding potentials

This work evaluates a maximum likelihood parameter estimation method for regions-of-interest (ML-ROI) when incorporated in a direct 4D PET image reconstruction framework including the simplified reference tissue model with the basis function method (SRTM-BFM) tracer kinetic model. The ML-ROI algorithm has been evaluated for the usual task of estimating the radioactivity concentration for ROI spatial-bases compared to voxels. We therefore extend the application of this method to include the direct estimation of binding potential (BP) values on simulated 2D+time data sets (with use of [11C]raclopride time-activity curves (TACs) from real data). The performance of the proposed method is evaluated by comparing BP estimates with those obtained from a conventional post reconstruction approach, as well as the original ML-ROI method. It is shown that the use of ROIs as spatial basis functions leads to much lower %RMSE for BP regional estimates (%RMSE reduced by a factor of 2 or more), and furthermore using direct BP estimation in conjunction with ROI spatial basis functions reduces the still further. However, the major improvement is from the use of ROI spatial basis functions, rather than the use of direct kinetic parameter estimation. On the other hand, the considerable time gained (2 orders of magnitude) makes it a potential candidate for routine application.