3D Regularisation and Segmentation of Factor Volumes to Process PET H2 15O Myocardial Perfusion Studies

A 3D image processing methodto increase reproducibility in the handling of PET myocardial perfusion studies is proposed. It is basedon the 3D regularization of factor volumes, which are estimated by FAMIS (Factor Analysis of Medical Image Sequences). The resulting regularized factor volumes correspond to right and left cavities and to perfused tissues. They are then submitted to a C-means classification, looking for 4 clusters. Some rules of connectivity are applied to the resulting clusters to achieve the segmentation of heart cavities and myocardium. Thereafter, kinetics in the left ventricle and in the myocardium can be computed from the previously determined volumes of interest, and myocardial blood flow (MBF) is estimated by a conventional compartmental analysis.