Improved parametric images of blood flow and vascular volume by: Cluster analysis in H₂¹⁵O brain PET study

The values of cerebral blood flow (CBF) and vascular volume (V₀) can be estimated using H₂¹⁵O dynamic PET and the two-compartment model. In this study, we present a method that can generate parametric images of both CBF and V₀, with improvement of image quality, by a single computational procedure. This method is based on the two-compartment model, and employs linear least square algorithm and cluster analysis for parameter estimation and suppressing noise on image data, respectively. The results in computer simulation showed that this method could provide the reduction of error in parameter estimation, as well as noise on parametric images of both CBF and V₀, and the smaller effect of changes of CBF and V₀ on the estimation of both parameters. In the PET study, this method could provide the images of CBF and V₀ with improvement in quality, compared with those without clustering by showing the clear location of arterial vascular components on the V₀ image. In the simulation, the error in parameter estimation was sufficiently small for K₁, but not for V₀. These results reveal that the presented method has the potential to make a contribution to the improved diagnosis of cerebral vascular disease and activation.