A STUDY OF PARTIAL VOLUME EFFECTS ON CLUSTERING-AIDED PARAMETRIC IMAGES

Cluster analysis has been applied in functional imaging to derive parametric images with improved signal to noise ratio. However, the impact of partial volume effects (PVE) on the parametric images aided by clustering is still uncertain. Computer simulations were performed to generate simulated dynamic data at various noise levels. Reconstructed data were processed by cluster analysis. The generalized linear least squares (GLLS) method was used to estimate parametric images. The results were compared with the parametric images generated without clustering. The results demonstrate that clustering does not enhance PVE, but reduces PVE. Furthermore, clustering-aided parametric images are shown to be insensitive to noise.