Iterative Automatic Segmentation in cardiac PET based on TAC correlation: preliminary results

Abstract

Conventional kinetic parameter estimation based on compartmental models requires an accurate estimation of arterial blood input function. To avoid invasive blood sampling, an image-derived input function can be obtained by manually defining a Region of Interest. Here we propose a new and simple, iterative method for automatic segmentation and input function calculation of PET cardiac studies using correlation as a distance metric between a priori information regarding the approximate shape of the final time-activity curve (TAC) and the actual TAC extracted from the image temporal series.