Improving signal-to-noise ratio in FDG parametric images by cluster analysis

The problems associated with forming parametric images from a positron emission tomography (PET) kinetic study are large noise levels in the voxel-based tissue time-activity curve (tTAC), and huge calculation times. A clustering based kinetics method is proposed for addressing these problems. In this method, voxels with the same shape are gathered and averaged before parameter estimation. Compared with an ordinal regions of interest (ROI) based non-linear parameter estimation method, parametric images with almost the same statistical reliability are obtained. In addition, the calculation time is very short (two minutes).