Edge detection in ultrasound images based on differential tissue attenuation rates.

Abstract

Identification of the cardiac borders is a basic goal in the analysis of echocardiographic images. This is a requirement for various analysis purposes, such as the derivation of quantitative parameters, for wall motion analysis and for the creation of three-dimensional reconstructions. Border identification is usually accomplished by first identifying a set of generic edge points, then extracting the border of interest from the edge map. A variety of edge detection algorithms have been reported in the literature. However, the most commonly used methods are all based on the same underlying tissue discriminant, namely, the widely differing backscatter levels of the different types of tissue that form the cardiac borders, in particular, blood and myocardium. We have implemented a new method of edge detection based on an entirely different ultrasonic tissue characteristic — that of attenuation, the rate at which the tissue absorbs ultrasonic energy. This characteristic also differs widely between blood and myocardium, and therefore provides an alternative means of discriminating between these two types of tissue. The method searches for the abrupt change in attenuation rate that characteristically occurs at the blood/myocardium interface. At each point in the image, the average attenuation rates of the portions of tissue radially preceding and radially following the point are determined by least-squares regression. If one of these attenuation rates is sufficiently close to that of blood, and the other is sufficiently close to that of myocardium, then the point is flagged as an edge point. This paper describes how this method is implemented in software, and presents the results of applying the method to a library of sample images.