Imaging glucose metabolism of brain using PET with clustering analysis for kinetics

The aim of this study is to visualize glucose metabolism in the brain using PET dynamic data and the proposed clustering analysis (CAKS). PET can give the glucose metabolism using /sup 18/F-FDG. Voxel-based analysis is not practical because of the bad noise property in voxel-based PET data and a large number of voxels. In CAKS, PET data are clustered to improve reliability in estimation and calculation speed. In CAKS, voxels whose concentration history in tissue time activity are similar, are gathered before parameter estimation using a statistical clustering algorithm based on the mixture Gaussian model, then the averaged time course is used for parameter estimation. As a result, physiologically acceptable images on the glucose metabolism were obtained in ten minutes.