IEEE Trans Ultrason. 2026 Sep 21;PP. doi: 10.1109/TUSON.2026.3735403. Online ahead of print.
ABSTRACT
Accurate cardiovascular flow quantification is required for diagnosing and managing valvular disease, heart failure, and atherosclerosis, and for assessing wall shear stress. Conventional Eulerian ultrasound vector flow imaging methods, such as echo particle image velocimetry (echoPIV), suffer from spatial averaging and reduced accuracy in boundary and high-gradient regions. We propose a Kalman filter (KF)-based echo-particle tracking velocimetry (echoPTV) framework that integrates motion prediction and Probabilistic Data Association (PDA) to enhance track continuity and robustness in fast, complex cardiovascular flows. In-silico simulations (jet, carotid, and left ventricle phantoms) demonstrated that KF-based echoPTV produced smoother, longer, and more coherent trajectories than Hungarian tracking, with significantly lower velocity bias (3-7% vs. 5-10%), reduced variability, and improvements of up to 42% in track duration similarity index and 5-13% in F1 score. In an in-vitro flow phantom, echoPTV recovered the 0.67 m/s peak velocity and achieved lower root-mean-square error relative to the reference parabolic profile (2 cm/s) compared to echoPIV (6-7 cm/s). In-vivo results revealed similar flow patterns between echoPIV and echoPTV, but higher peak inflow velocities with echoPTV. These results demonstrate the potential of KF-PDA echoPTV for accurately capturing complex, high-velocity cardiovascular flows.
PMID:42776886 | DOI:10.1109/TUSON.2026.3735403