3.5D dynamic PET image reconstruction incorporating kinetics-based clusters.

摘要: Standard 3D dynamic positron emission tomographic (PET) imaging consists of independent image reconstructions individual frames followed by application appropriate kinetic model to the time activity curves at voxel or region-of-interest (ROI). The emerging field 4D PET reconstruction, contrast, seeks move beyond this scheme and incorporate information from multiple within reconstruction task. Here we propose a novel framework aiming enhance quantitative accuracy parametric images via introduction priors based on kinetics, as generated clustering preliminary reconstructed define clustered neighborhoods voxels with similar kinetics. This is then straightforward maximum posteriori (MAP) applied frames; such method labeled ‘3.5D’ reconstruction. use cluster-based has advantage further enhancing performance in imaging, because: (a) there are typically more clusters than conventional local neighborhoods, (b) neighboring distinct kinetics less likely be together. Using realistic simulated 11C-raclopride data, proposed was investigated. Parametric distribution-volume (DV) DV ratio (DVR) were estimated using maximum-likelihood expectation maximization (MLEM), MAP quadratic prior (QP-MAP), (c) Green (GP-MAP) (d, e) two (CP-U-MAP CP-W-MAP), graphical modeling, qualitatively quantitatively compared for 11 ROIs. Overall, methodology resulted substantial visual well improvements (in terms noise versus bias performance) DVR images. also tested 90 min patient study performed high-resolution research tomography. shown outperform regional value performance).