Quantitative comparison of new image processing methods for volumetric analysis of left ventricular contrast echocardiograms

Abstract

An effort has been made to develop image processing methods which allow a definite and precise tracking of the borderline of the ventricle in two-dimensional echocardiograms. Experience is reported with two new methods, which are based on the gray-level rise (GL) and the signal-to-noise ratio (SNR) in combined heart-phase-triggered image series. A quantitative comparison of these time-series methods is presented with respect to the interpretation of a single native image (noncontrast image), a single contrast-material image, the corresponding subtraction image, and the corresponding color superposition image. The comparison is based on the calculation of the ejection fraction using the above-mentioned methods, interobserver variability, and intraobserver variability. The color superposition method and the two new methods described here (GL and SNR) are shown to be especially well suited for left-ventricle function evaluation.<>