Three-dimensional echocardiography: the influence of number of component images on accuracy of left ventricular volume quantitation.

ABSTRACT

One approach to three-dimensional echocardiography is to reconstruct the surface of cardiac structures from two-dimensional images positioned in three-dimensional space. This approach has yielded accurate measures; however, the relationship between the number of nonparallel images used in three-dimensional echocardiographic reconstruction to the accuracy of the volume calculated has not been determined. With a canine model in which instantaneous left ventricular volume could be measured in vivo, images were obtained from intersecting long- and short-axis scans and stored with their spatial coordinates. The left ventricle was reconstructed and its volume calculated. The difference between three-dimensional echocardiographic and true volume was determined in 84 different cavitary volumes (4 to 85 ml). In each case, long- and short-axis images were deleted serially from the original data set (maximum of 27) until there were only three images left in the reconstruction. After each set of deletions, left ventricular volume was recalculated with the remaining images. Three-dimensional echocardiography accurately quantified ventricular volume with eight to 12 intersecting images, with a mean error of less than 1 ml and an SD of 5 ml. With a reduction of component images below eight, there were progressive increases in both absolute and mean percentage error. Accurate assessment of stroke volume and ejection fraction in this beating heart model also required eight to 12 images. Left ventricular volume and systolic function can be quantitated by three-dimensional echocardiography with as few as eight to 12 intersecting or nonparallel images.