Automated Measurement of Local Mechanical Activation on High Frame Rate Echocardiography

Abstract

High Frame Rate Speckle Tracking Echocardiography (HFR-STE) offers a method to pinpoint the local onset of contraction in the left ventricle (LV) and generate mechanical activation maps. In this paper, a new patient-specific spatiotemporal approach is proposed to identify activation times on left ventricular strain rate (SR) curves automatically. Curves are collected from 2D HFR-STE according to the 16-segment model. Using a Locally Weighted Principal Component Analysis (LWPCA), the main pattern of each segment’s SR curve is extracted locally. The first positive-to-negative zero-crossing point on the first principal component is identified as the activation time. Validation with a dataset of 40 subjects (20 healthy volunteers and 20 patients) showed that 94% of estimated activation times closely matched the expert-identified times, differing by no more than 16ms. Quantitative and qualitative comparisons between LWPCA and (weighted) averaging are also reported. Also, the automatically generated activation maps closely resemble their manually created counterparts, demonstrating good visual similarity.