Model-based flow quantification using conditional random fields with global preferences

Recent advances in the real-time 3D full volume echocardiography have enabled high frame rate acquisition of volumetric color Doppler flow images. In this paper, we propose a fully automated method for de-aliasing of color Doppler flow and quantification of the cardiac flow using instantaneous 3D+t ultrasound data. First, the anatomical information, such as mitral annulus and left ventricle outflow tract (LVOT), are detected and tracked automatically accounting for the heart motion. 3D planes are constructed at the estimated mitral annulus and LVOT locations to sample the color flow data. To overcome inherent limitation of the color Doppler velocity ambiguity caused by the Nyquist velocity level, we formulate the de-aliasing process in a conditional random field (CRF) framework and introduce a global consistency potential to incorporate domain specific prior information based on the instantaneous change in LV volume. Preliminary results with clinical data presented here agree well with clinical measurements in a quantitative manner. The proposed method is efficient and achieves high speed performance of 0.2 second per volume of ultrasound data.