Improved Signal-to-Noise Ratio in Parametric Images by Cluster Analysis☆

摘要: Abstract Parametric images are formed by analyzing the concentration history of every voxel in PET data sets. Because at level rather noisy, noise propagation into parametric image is often quite noticeable. To address this problem, a model-based clustering method has been developed to generate images. The basic idea average over voxels whose histories have same shape. We applied two-parameter ( K 1 ,  k 2 ) compartment model local cerebral blood flow. statistic R = ∫  tC t )  dt /∫  C te − ⊗ e classifies curves terms , where and denote tissue histories, respectively, convolution operator. Simulation studies showed that 30% yielded 2% standard deviation . flow partition coefficient were computed for an O 15 study, with without clustering. Cluster size affected bias, statistical precision, computation time. With clusters 400 voxels, variance parameter was around 1/50 smaller clustering, negligible bias time 30 s on 64-MHz workstation × 128 MATLAB 5.1.