Development and Validation of a Phase-Filtered Moving Ensemble Correlation for Echocardiographic Particle Image Velocimetry

Abstract

A new processing method for echocardiographic particle image velocimetry (EchoPIV) using moving ensemble (ME) correlation with dynamic phase correlation filtering was developed to improve velocity measurement accuracy for routine clinical evaluation of cardiac function. The proposed method was tested using computationally generated echocardiogram images. Error analysis indicated that ME EchoPIV yields a twofold improvement in bias and random error over the current standard correlation method (β_Pairwise = -0.15 vs. β_ME = -0.06; σ_Pairwise = 1.00 vs. σ_ME = 0.49). Subsequently a cohort of eight patients with impaired diastolic filling underwent similar evaluation. Comparison of patient EchoPIV velocity time series with corresponding color M-mode velocity time series revealed better agreement for ME EchoPIV compared with standard PIV processing (R_ME = 0.90 vs. R_Pairwise = 0.70). Further time series analysis was performed to measure filling propagation velocity and 1-D intraventricular pressure gradients. Comparison against CMM values indicated that both measurements are completely decorrelated for pairwise processing (R²_Vp = 0.15, R²_IVPD = 0.07), whereas ME processing correlates decently (R²_Vp = 0.69, R²_IVPD = 0.69). This new approach enables more robust processing of routine clinical scans and can increase the utility of EchoPIV for the assessment of left ventricular function.

Keywords: Diastolic dysfunction; Echocardiography; Particle image velocimetry.