Sensitivity and specificity of left ventricular ejection fraction by echocardiographic automated border detection: Comparison with radionuclide ventriculography

Abstract

BACKGROUND: Echocardiographic automated border detection (ABD) provides on-line, beat-to-beat estimation of left ventricular (LV) ejection fraction (EF). Sensitivity and specificity of using ABD-EF for diagnosing LV dysfunction in routine clinical situations have not been previously studied. HYPOTHESIS: Analysis of ABD-EF data based on receiver operating characteristic (ROC) should provide useful information about sensitivity and specificity for clinical diagnosis of LV function based on ABD-EF. METHODS: The study group included 50 consecutive patients with EF measured by both ABD and radionuclide ventriculography (RVG). ABD-EF was recorded for 25 consecutive heart beats in the apical four-chamber view. Data were analyzed statistically by linear regression, Bland-Altman plot, and ROC. In ROC analysis, abnormal LV function was defined RVG-EF < or = 40%. RESULTS: ABD and RVG showed a moderate correlation in the EF measurements: slope = 0.93, intercept = 17%, r = 0.79 (n = 50). Interbeat variability in ABD was diminished by averaging consecutive beats; standard error of estimate (SEE) decreased from 15.6% without averaging to 12.5% with 25-beat averaging. Bland-Altman analysis indicated that ABD-EF compared unfavorably with RVG-EF, with limits of agreement from -11% to 39%. ABD-EF showed a systematic overestimation (p < 0.005), which was compensated by increasing the threshold for abnormal ABD-EF to 56%. With the optimized threshold, ABD-EF provided 89% sensitivity and 89% specificity (85% overall diagnostic accuracy) for diagnosing abnormal LV function. CONCLUSION: This study explored the limitations of on-line echocardiographic measurement of EF in a clinical setting and provided useful data for assessing interbeat variability, sensitivity, and specificity.