Learning-based 3D myocardial motion flowestimation using high frame rate volumetric ultrasound data

The estimation and analysis of cardiac motion provides important information for the quantification of the elasticity and contractility of the myocardium. Taking advantage of the recent progress on real-time ultrasound imaging, unstitched volumetric data can be captured in a high frame rate. In this paper, we propose a learning-based method to automatically estimate the 3D displacements and velocities of the myocardial motion. To achieve robust tracking on ultrasound image sequences, multiple information is fused together in our framework to handle noisy and missing data, including speckle patterns, boundary detection and motion prediction. Preliminary results on clinical data confirmed these findings in a qualitative manner. The estimated displacement and velocity values have a strong agreement with the results from other systems and modalities. The proposed method is efficient and achieves high speed performance of less than 1 second per frame for volumetric ultrasound data.