Effects of noise on the determination of ejection fraction from left ventricular time-activity curves

The effects of Poisson noise on three estimates of ejection fraction made from left-ventricular time-activity curves have been investigated. All three methods are based on a sinusoidal model of left-ventricular volume changes. The first, developed by Schelbert et al., overestimates the ejection fraction for low-activity levels and low ejection fractions. The second estimate, which is merely a first-order correction for the contribution of Poisson noise to the first estimate, appears to be more accurate when both estimators are applied to simulated time-activity curves, and the resulting ejection fractions are compared. A third, "maximum likelihood" estimator, when applied to the same data, is apparently more accurate than the first two.