Background: The accuracy of left ventricular (LV) volumes and ejection fraction (EF) on two-dimensional echocardiography (2DE) is limited by image position (IP), geometric assumption (GA), and boundary tracing (BT) errors.
Methods: Real-time three-dimensional echocardiography (RT3DE) and cardiac magnetic resonance imaging (CMR) were used to determine the relative contribution of each error source in normal controls (n = 35) and patients with myocardial infarctions (MIs) (n = 34). LV volumes and EFs were calculated using (1) apical biplane disk summation on 2DE (IP + GA + BT errors), (2) biplane disk summation on RT3DE (GA + BT errors), (3) 4-multiplane to 8-multiplane surface approximation on RT3DE (GA + BT errors), (4) voxel-based surface approximation on RT3DE (BT error alone) and (5) CMR. By comparing each method with CMR, the absolute and relative contributions of each error source were determined.
Results: IP error predominated in LV volume quantification on 2DE in normal controls, whereas GA error predominated in patients with MIs. Underestimation of volumes on 2DE was overcome by increasing the number of imaging planes on RT3DE. Although 4 equidistant image planes were acceptable, the best results were achieved with voxel-based RT3DE. For EF estimation, IP error predominated in normal controls, whereas BT error predominated in patients with MIs. Nevertheless, one third of the EF estimation error in patients with MIs was due to a combination of IP and GA errors, both of which may be addressed using RT3DE.
Conclusions: The relative contribution of each source of LV quantitation error on 2DE was defined and quantified. Each source of error differed depending on patient characteristics and LV geometry.