Abstract 3143: Sources of Error for Left Ventricular Quantitation Differ Among Patients with Normal and Post-Myocardial Infarction Ventricles: Insights from Real-Time 3D Echocardiography and Cardiac Magnetic Resonance Imaging
Accuracy of LV volumes and EF by 2DE is limited by errors due to 1) uncertain image position so that there is inadequate coverage of the LV (apical two and four-chamber views are assumed to be orthogonal, but image position is not verified), 2) due to incorrect assumptions of geometric shape (ellipsoid), and 3) incorrect boundary recognition. However, the relative contribution of each may be different among normal and post-MI ventricles. We used 3D echo (3DE) and cardiac magnetic resonance imaging (CMR) to quantify these sources of error.
Methods: LV volumes were calculated in 23 normal subjects and 23 post-MI patients using:
2DE apical biplane summation of disks,
3DE biplane summation of disks,
3DE surface approximation from 10 rotationally equidistant apical slices (TomTec),
CMR using contiguous, short axis, TrueFISP cine images.
Image position error is associated with method (a). Geometric assumption error is associated with methods (a) and (b). Boundary recognition error is noted for methods (a), (b) and (c) relative to CMR.
Results: EDV, ESV and EF obtained by 2DE and 3DE were compared to CMR by Pearson’s correlation, simple linear regression, Bland-Altman analysis and ANOVA. Absolute and relative root mean square percent (RMS %) error were calculated using CMR as a reference (table⇓). ANOVA shows that for normal and post-MI subjects, LV volumes obtained by the 2DE and 3DE biplane methods were significantly different from CMR, whereas 3DE-10 plane did not differ from CMR. For EF calculation in abnormals, however, both 3DE biplane and 3DE-10 plane methods did not differ from CMR.
Conclusions: Relative RMS% error values show that whereas image position error is the primary source of error in quantfying 2DE LV volumes in normals, geometric assumption error predominates in post-MI patients. 2DE EF results show that two-thirds of error can be attributed to residual boundary recognition error which may be addressed with administration of intravenous contrast agents.