Quantitative comparison of LV volumes in 3D echocardiography and cardiac magnetic resonance imaging

Objective: To study the different endocardial visualization in three-dimensional echocardiography (3D echo) and cardiac magnetic resonance (CMR) by aligning 3D echo and CMR image slices in three dimensions. Methods Six subjects (4 with myocardial infarction) underwent 3D echo and CMR (short- and long-axis cine 2D). In-house developed software was used to align the CMR short- and long-axis slices to compensate for different breath-hold positions. Alignment of the 3D echo and CMR datasets was based on a set of 7 landmarks (apex and six mitral valve attachment points) identified interactively by an experienced cardiologist, both in the 3D echo and CMR data. A least square fitting scheme was employed to achieve optimal 3D alignment. Results 3D echo underestimated LV volumes and ejection fraction (EF) compared to CMR (end-diastolic (ED) 50±32 ml, end-systolic (ES) 17±39 ml, EF 10±5%). However, the correlations were high: ED r=0.98, ES r=0.99, EF r=0.99. The average difference (mm) between the drawn contours from 3D echo (red) and CMR (green) was lower in the antero- and inferoseptal segments (Fig. 1: ED contours for both methods overlaid on 4-, 2- chamber 3D echo (left), bulls eye plot of the average distance between them (right)). Conclusions As expected, 3D echo consistently underestimates LV volumes, and we suggest that this difference is mainly caused by different endocardial visualization in the lateral and inferior walls. This is vital information in studies involving both modalities, and in the development of new automatic 3D echo methods for volume measurements.