Digital image processing of two dimensional echocardiograms: identification of the endocardium

Digital computer image-processing techniques were applied to two dimensional echocardiograms to improve the accuracy of cardiac spatial measurements by enhancing endocardial recognition. Images were photographed from the two dimensional echocardiographic monitor and digitized using an optical densitometer. Image-processing algorithms were applied to the digitized images as follows: (1) Multiple images were averaged; (2) a gray level threshold was chosen to separate the image into tissue and cavity regions on the basis of amplitude (brightness) of the returning echoes; (3) endocardium was traced between the regions; (4) endocardial position was confirmed by matching this boundary with a contrast edge map of the original images; and (5) the endocardial boundaries were tested by comparison with simultaneous M mode echocardiograms. A linear correlation was found between M mode and computer-processed two dimensional echocardiographic measurements of ventricular septal thickness (r = 0.88); this was superior to the correlation between M mode and unprocessed two dimensional echocardiographic septal measurements (r = 0.55). The correlations between M mode and processed or unprocessed two dimensional echocardiographic measurements of left ventricular internal dimension were similar (r = 0.89 and 0.85, respectively), but the slope of the regression line for the processed data more closely approximated the line of identity (p less than 0.05). It is concluded that endocardial outlines derived with use of digital image-processing techniques lead to left ventricular measurements that correlate more closely with M mode measurements than do dimensions derived from unprocessed two dimensional echocardiography.