Use of anatomical information in a Bayesian reconstruction with an edge-preserving median prior

Abstract

We have previously proposed a maximum a posteriori (MAP) reconstruction with a median prior using convergent ordered subsets expectation maximum algorithm, called MAPCOSEM-MP. In contrast to the smoothing prior that imposes global smoothness, the median prior enhances edges and simultaneously retains local smoothness. Herein, we use simulations to investigate whether the incorporation of anatomical information in MAPCOSEM-MP can provide more accurate quantitation. The simulation results show that the introduction of anatomical information in the MAPCOSEM-MP reconstruction can further improve the quantitation as well as the image quality. Moreover, we find that the anatomy-based MAPCOSEM-MP reconstruction is less sensitive to registration errors of 1 to 2 pixels between functional and anatomical images. This finding may indicate that MAPCOSEM-MP has the ability to reduce artifacts caused by inaccurate registration. We expect that the improved performance in quantitation could provide better image quality for disease detection.