Simultaneous tracking of endocard and epicard in ultrasonic contrast imaging

Knowing the geometry of both endocard and epicard is important in both qualitative and quantitative evaluation of the cardiac function. The reliability of existing algorithms for automatic wall tracking in 2D echocardiography decreases as the signal-to-noise ratio decreases. The local edge definition can be missing for both endocard and epicard. These image dropouts can be filled in by localizing the contour with available edge information and adjusting the other contour by requiring stability in the wall thickness. This approach requires a simultaneous extraction of both endocard and epicard. In this paper we accomplish this by using a generalized computational framework of active contours, and demonstrate initially the superiority of the approach in both synthetic and real ultrasound images. This has been implemented by modeling legal configurations for both endocard and epicard. The local energy contributions can hence be defined freely as a function of contour curvature, wall thickness, derivative of wall thickness, and image signatures at both endocard and epicard