Quantitative texture analysis in two-dimensional echocardiography: application to the diagnosis of experimental myocardial contusion.

Abstract

We postulated that the analysis of regional image texture in two-dimensional (2D) echocardiograms would be an accurate method to differentiate normal from abnormal myocardial structure. We tested this hypothesis with quantitative texture measures to study the regional, spatial distribution of echo amplitudes in 2D echocardiograms performed before and immediately after blunt left chest trauma was induced in six anesthetized dogs. After trauma the contused region of myocardium appeared brighter and exhibited an altered myocardial texture. By use of a set quantitative texture measures, we found no significant differences in pretrauma images when normal regions were compared with regions to be contused. Also, we found no difference when we compared the normal regions in each animal in pretrauma vs posttrauma images. Twelve measures, however, differentiated normal from contused regions within the posttrauma images (p values ranged from .0057 to .0001 by multivariate analysis of variance). These texture measures were capable of differentiating normal from abnormal tissue only when texture along the azimuthal (lateral) direction was calculated. We conclude that regions of myocardial contusion exhibit visibly altered local echo-amplitude patterns (altered image texture) and that these image texture alterations may be quantified with digital image analysis techniques. These findings suggest that quantitative texture calculations may be a useful approach to ultrasound tissue characterization.