The Nuclear Fusion Award

[摘要] In this study, we aim to reconstruct single-photon emission computedtomography images using anatomical information from magnetic resonanceimaging as a priori knowledge about the activity distribution. The trade-offbetween anatomical and emission data is one of the main concerns for suchstudies. In this work, we propose an anatomically driven anisotropic diffusionfilter (ADADF) as a penalized maximum likelihood expectation maximizationoptimization framework. The ADADF method has improved edge-preservingdenoising characteristics compared to other smoothing penalty terms based onquadratic and non-quadratic functions. The proposed method has an importantability to retain information which is absent in the anatomy. To make ourapproach more stable to the noise-edge classification problem, robust statisticshave been employed. Comparison of the ADADF method is performed witha successful anatomically driven technique, namely, the Bowsher prior (BP).Quantitative assessment using simulated and clinical neuroreceptor volumetricdata show the advantage of the ADADF over the BP. For the modelled data, theoverall image resolution, the contrast, the signal-to-noise ratio and the ability topreserve important features in the data are all improved by using the proposedmethod. For clinical data, the contrast in the region of interest is significantlyimproved using the ADADF compared to the BP, while successfully eliminatingnoise.