Fully Automated Software for Polar-Map Registration and Sampling from PET Images

Left ventricle myocardial activity (LV) from volumetric images is commonly presented in a 2D format called a polar map. Few applications offer complete automation of the registration and sampling process, which may lead to biasing of the results by the user and increased variability. A fully automated process is described which includes two steps: reorientation of the data to a common LV reference frame, and registration of the LV using a deformable spline model. The software implementation of this algorithm consists of a graphical user interface (GUI) that enables the user to verify the process, and to intervene in cases where automation may fail. 135 PET images (consisting of several tracers, in rest and stress, and in humans and dogs) were processed and visually validated. All but one case (99.3%) were successfully handled by the automated process.