Automated 3D Velocity Estimation of Natural Mechanical Wave Propagation in the Myocardium

Abstract

The mechanical wave (MW) propagation velocity in the heart is related to the tissue stiffness and its measurement mainly relies on manual evaluation of the 1D wave projection. This study presents an automated method for 3D wave visualization and velocity estimation in the heart using 3D ultrasound imaging of the left ventricle (LV). High-quality (HQ, 19 vps) and high-frame-rate (HFR, 823 vps) volumes were acquired. Deep learning models automatically segmented the LV and extracted the apical standard views from the HQ data which were used to derive the anatomical M-lines and myocardial segmentation. The clutter filter wave imaging (CFWI) and tissue Doppler imaging (TDI) generated wave propagation maps from HFR data, and the aortic valve closure (AVC) and atrial contraction/kick (AK) waves were automatically detected. LV segmentation and anatomical M-lines were used for 3D wave propagation extraction and its 1D projection, respectively. The 1D wave propagation velocity was determined through automatic slope detection, while the 3D velocity map was derived from the gradient of the time-of-flight (TOF) map. Results showed varying 1D velocity across views and myocardial regions, with the AVC propagation velocity surpassing that of the AK wave. The pipeline remained stable and generated results consistent with expert measurements. Comparing 3D and 1D propagation highlighted errors from 1D projection and demonstrated the benefits of the 3D method in assessing regional velocities and the validity of the 1D approach. This study demonstrated an automatic evaluation of 3D MW propagation velocities in the entire LV, leading to improved accuracy and standardized measurements of myocardial tissue properties.