Click on image to view larger version.

Figure 1 Method of semi-automatic contour detection using real-time three-dimensional echocardiography (3DE): a stack of eight rotated long-axis images is automatically selected from the 3DE dataset (left panel). In each of these long-axis cross sections (only one shown), the left ventricular (LV) apex and mitral annulus are earmarked in an end-diastolic (middle upper panel) and end-systolic (middle lower panel) stop frame. The markers at the mitral annulus are subsequently used by the algorithm to set the mitral valve plane and to truncate the contours of the LV cavity. Thereafter, an ellipse is placed in each of the end-diastolic and end-systolic stop frames for initiation of the model. This ellipse is manually adapted (length, width, and rotation angle) to fit as closely to the endocardial border as possible. The contours are subsequently detected automatically by the algorithm (right upper and lower panels for end-diastolic and end-systolic stop frames, respectively, with detected contour). See text for further details.