Real-time segmentation of 4D ultrasound by Active Geometric Functions

Abstract

Four-dimensional ultrasound based on matrix phased array transducers can capture the complex 4D cardiac motion in a complete and real-time fashion. However, the large amount of information residing in 4D ultrasound scans and novel applications under interventional settings pose a big challenge in efficiency for workflow and computer-aided diagnostic algorithms such as segmentation. In this context, a novel formulation framework of the minimal surface problem, called Active Geometric Functions (AGF), is proposed to reach truly real-time performance in segmenting 4D ultrasound data. A specific instance of AGF based on finite element modeling and Hermite surface descriptors was implemented and evaluated on 35 4D ultrasound data sets with a total of 425 time frames. Quantitative comparison to manual tracing showed that the proposed method provides LV contours close to manual segmentation and that the discrepancy was comparable to inter-observer tracing variability. The ability of such realtime segmentation will not only facilitate the diagnoses and workflow, but also enables novel applications such as interventional guidance and interactive image acquisition with online segmentation.