Application of factor and cluster analysis for the parametric image of myocardial blood flow using H/sub 2//sup 15/O and dynamic PET

Although the parametric image (PI) of regional myocardial blood flow (rMBF) using H/sub 2//sup 15/O and PET would facilitate the numerical evaluation and the clinical visual interpretation of rMBF, low signal to noise ratio (SNR); of the time activity curve (TAC) of individual voxel in dynamic H/sub 2//sup 15/O PET image has made the composition of PI difficult to accomplish. The authors propose a method for PI of rMBF from H/sub 2//sup 15/O PET using factor analysis and cluster analysis. Factor analysis was performed to extract TACs of the right and left ventricular blood pool and their factor images. The blood pool activity was then subtracted from the original dynamic frames to remove the spillover contamination to myocardial tissue. Pure TACs in myocardial tissue were submitted to cluster analysis for PI of rMBF, which averaged voxels with the same shape TACs to increase SNR of PI and reduce calculation time. Parametric images showed good image quality and contrast in comparison to either the factor images of myocardial tissue or the subtracted imaged composed by subtracting the initial 30 seconds' images from the 2 minutes' ones. RMBF ranged 0.89-1.35 m/min/g at rest, and 1.16-3.85 at stress. Computation time was less than 1 minute on the workstation with 333 MHz CPU.