Automatic TAC extraction from dynamic cardiac PET imaging using iterative correlation from a population template

Abstract

This work describes a new iterative method for extracting time-activity curves (TAC) from dynamic imaging studies using a priori information from generic models obtained from TAC templates. Analytical expressions of the TAC templates were derived from TACs obtained by manual segmentation of three (13)NH3 pig studies (gold standard). An iterative method for extracting both ventricular and myocardial TACs using models of the curves obtained as an initial template was then implemented and tested. These TACs were extracted from masked and unmasked images; masking was applied to remove the lungs and surrounding non-relevant structures. The resulting TACs were then compared with TACs obtained manually; the results of kinetic analysis were also compared. Extraction of TACs for each region was sensitive to the presence of other organs (e.g., lungs) in the image. Masking the volume of interest noticeably reduces error. The proposed method yields good results in terms of TAC definition and kinetic parameter estimation, even when the initial TAC templates do not accurately match specific tracer kinetics.

Keywords: Automatic segmentation; Cardiac imaging; Kinetic modeling; Pet.