Extraction and validation of parameters from two-dimensional echocardiographic images

The potential for making quantitative parameter measurements from two-dimensional computer reformatted ECG gated echocardiographic images has been evaluated in a comparison study with cardiac angiography. The parameters considered in this study are left ventricular long and short axis, volume, ejection fraction, and regional percent shortening. A theoretical analysis of the potential sources of uncertainty in measurements derived from echocardiographic images is given. The error sources included are physical dimensions of the ultrasound pulse, uncertainty in the velocity of sound, reproducibility of border input, inaccuracies in deriving three-dimensional measurements from two-dimensional images, and those uncertainties associated with the choice of scanning plane and movements of the ventricle perpendicular to this plane. The standard errors observed between sets of measurements made using the echocardiographic images and corresponding RAO left ventricular angiograms were: 33 ml or 22 percent for end-diastolic volumes, 15 ml or 30 percent fro end-systolic volumes, 10 mn or 10 percent for long axis length, and .11 or 17 percent for ejection fraction. Regional percent shortening was measured using hemi-diameter, radial, and elliptical models on contours drawn on echocardiographic and angulated half-axial LAO angiographic images, this later view closely corresponding to that of the echocardiographic images. The observed standard error is approximately 50 percent of the actual percent shorting. For all measurements except ejection fraction the observed standard errors were considerably greater than expected. This is attributed in part to the fact that the angiographic and ultrasonic measurements were not made simultaneously and that the angiographic technique is more traumatic to the patient. For this reason the physiological state of the patient was not comparable in all cases between the two measurement techniques. Other error sources include uncertainties in structure recognition in both echocardiography and angiography, and the inability of echocardiography to visualize the apex of the left ventricle. The fact that the comparison of ejection fractions results in a slightly lower than expected standard error indicates that much of the error in measurement on end-diastolic and end-systolic images from the same patient are of similar magnitude and direction. Because the image is degraded due to the finite dimensions of the ultrasound wave packet, a lateral filter to enhance the image has been developed. By applying the lateral filter to ECG gated echocardiographic images it is observed that structure continuity is enhanced an low level, high frequency echoes which are probably due to structures our of the scanning plane are considerably reduce in amplitude.