Three-dimensional (3D) transesophageal echocardiography (TEE) technology is now widely used intraoperatively in cardiac surgery. Left ventricular (LV) measurements with 3D transthoracic echocardiography correlate better with cardiac magnetic resonance measurements compared with traditional two-dimensional (2D) transthoracic echocardiography. In this study, we compared intraoperative 3D TEE against 2D TEE regarding quantitative indices of LV function.
We performed 2D TEE and 3D TEE examinations on 156 patients scheduled for elective cardiac surgery. Two-dimensional TEE images of midesophageal 4-, 2-chamber, and long-axis views were acquired. LV volumes and ejection fraction (EF) were calculated by Simpson's method. Three-dimensional full-volume images were recorded to calculate by a semiautomated procedure LV volumes (indexed to body surface area) and EF. 3D and 2D LV dimensions and function, image quality, time for acquisition/analyses, and reproducibility were compared by the Wilcoxon matched-pairs signed-ranks test. Pairwise differences between 3D and 2D data were compared using 95% prediction intervals (PIs) and Bland-Altman methodology. 3D volumes were also plotted against 2D volumes in scatter plots using a 3-zone error grid.
There was no significant difference between 3D and 2D in the estimation of EF (P = 0.227; median pairwise difference, -0.4% [95% PIs, -8.6% to 8.8%]). 3D LV indexed end-diastolic volumes (iEDVs) and end-systolic volumes (iESVs) were larger than 2D iEDVs (P < 0.001; median pairwise difference, 3.3 mL/m [95% PIs, -9.4 to 14.1 mL/m] and iESV: P < 0.001; median pairwise difference, 1.4 mL/m [95% PIs, -5.2 to 10.1 mL/m]). In the vast majority of cases (98.8% of cases for iEDV and 92.8% of cases for iESV), the difference between 2D and 3D TEE indexed volumes did not alter classification into normal, mildly to moderately dilated, or severely dilated volumes, as demonstrated by the 3-zone error grid analysis. Acquisition of 3D TEE image and analysis were not feasible in 4 patients (2.5%) for whom a quantitative 2D assessment of the LV was also impossible. 3D and 2D quality image was similar (P = 0.206). There was no difference in 3D versus 2D acquisition time (P = 0.805; pairwise difference = 2 seconds [95% PIs, -20 to 35 seconds]), but 3D analysis required more time (P < 0.001; pairwise difference = 117 seconds [95% PIs, 66 to 197 seconds]). Differences in repeated 3D versus 2D indexed volumes were not statistically significant, both considering interobserver reproducibility (iEDV: P = 0.125; pairwise difference, 0.26 ± 1.76 mL [95% PIs, -3.58 to 3.73 mL] and iESV: P = 0.126; pairwise difference, -0.16 ± 1.67 mL [95% PIs, -3.96 to 3.69 mL]) and intraobserver reproducibility (iEDV: P = 0.975; pairwise difference, -0.02 ± 1.20 mL [95% PIs, -2.32 to 2.08 mL] and iESV: P = 0.228; pairwise difference, -0.19 ± 1.13 mL [95% PIs, -2.47 to 2.53 mL]).
Intraoperative 3D TEE quantification of LV global function, image acquisition time, and reproducibility was not statistically different when compared with 2D TEE. It was however associated with calculation of larger LV volumes and a longer analysis time. Nevertheless, the 3-zone error grid analysis of the LV indexed volumes showed that the difference between 3D and 2D measurements does not affect the LV classification as normal, mildly to moderately dilated, or severely dilated.