An Extravascular Density Model for PET Partial Volume Correction

Abstract

Accurate quantification of /sup 18/F-FDG myocardial uptake values can be complicated by partial volume losses (PV). The use of an extravascular (EV) density image has been proposed as a means of correcting for PV losses in the left ventricle (LV) myocardium. The EV image, initially developed for lung density measurements, is created by subtracting a blood pool image from a transmission image. Within the myocardial wall, the EV values represent the mass of tissue present for tracer uptake. Increased PV losses correspond to decreased EV values. Division of the regional myocardial (MYO) values by the corresponding EV values should yield PV-free MYO estimates. As well as motion- and resolution-based PV losses, the EV value also removes the effects of the vascular component in the myocardial wall. In this paper, the EV model is developed mathematically and the assumptions of the model are discussed. Simulations performed to validate the model, indicated that the EV values could exactly estimate PV losses, under ideal conditions. However, a 2 mm difference between emission and transmission resolution was found to bias the EV value by /spl sim/5%. As well, transmission and activity spillover from organs adjacent to the LV myocardium violated the assumptions of the model leading to as much as a 15% underestimation of the corrected activity in the anterior and inferior walls. It was found that endocardial placement of the ROIs used to determine MYO and EV values minimized the effect of this spillover.