Two new echocardiographic methods for estimating left ventricular ejection fraction, both based on segmental wall motion scoring on a continuous scale, were developed and compared to radionuclide angiography in 50 patients. The first method used a geometric model in which the ventricle was represented by a tubular base section joined to a hemielliptical apex section, each containing 4 quadrants. Contraction was scored visually in 12 regions of the ventricle. The 12 scores were converted to dimensional shortenings, which were then used to calculate separate ejection fractions for each of the 8 quadrants. These 8 values were averaged to obtain the global ejection fraction. Assuming an elliptical end-systolic cross-section shape for each quadrant, the correlation between the echocardiographic and radionuclide ejection fractions was 0.917 with a regression slope of 0.93 and an intercept of 1.5. Results were identical assuming circular end-systolic quadrant shapes. The second approach used a linear combination of the mean wall motion score at the base, midleft ventricle and apex to predict ejection fraction. Coefficients for the 3 mean scores were optimized by least squares best fit with radionuclide ejection fraction in all 50 patients. The simplest linear model showed a correlation of 0.90 between the echocardiographic and radionuclide ejection fraction but underestimated low, and overestimated high, ejection fractions. However, these systematic errors could be removed by regression, which added a constant term of 8.1 and a slope factor of 0.87 to the linear model, improving the correlation with radionuclide ejection fraction to 0.91.(ABSTRACT TRUNCATED AT 250 WORDS)