Automated vs. Standard Tracking of LVEF and Longitudinal Strain
What is the feasibility and reproducibility of a fully automated system for measurement of left ventricular (LV) volumes, ejection fraction (EF), and longitudinal strain (LS)?
The FAST-EFs study investigators evaluated apical four-chamber (4C) and two-chamber (2C) images using commercially available ultrasound platforms from 255 patients in sinus rhythm at four centers. Local measurements included assessment of visual EF, determination of EF by biplane Simpson’s method. A single cycle in sinus rhythm, free of arrhythmia, was selected for each of the 255 patients. Images in DICOM format were then forwarded to a central laboratory for re-analysis of visual and Simpson’s rule EF, as well as determination of LS and determination of volumes and EF by the fully-automated system (AutoLV, TomTec-Arena 1.2, TomTec Imaging Systems, Unterschleissheim, Germany). The fully automated system was activated by noting that the image presented was a 4C or 2C view and required no manual interaction afterwards.
As the core laboratory was one of the study sites, the number of comparisons of visual, manual, and automated analyses data varied with the comparison in an effort to avoid possible observer bias. After identification of the apical 4C and 2C view, average analysis time was 8.8 ± 1 seconds/patient. Agreement between various measures was assessed with intraclass correlation coefficients (ICCs) and Bland-Altman analysis. The ICC for local center visual EF compared to local center manual EF, reference center visual EF, reference center manual EF, and automated EF were 0.82, 0.73, 0.74, and 0.70 with Bland-Altman bias of 2.7, -0.6, -2.2, and -2.5, respectively. The ICC for local center manual EF compared to reference center visual EF, reference center manual EF, and automated EF was 0.87, 0.84, and 0.83 with bias of 2.4, 0.9, and 0.2. There was 0 variability for automated EF and LS.
A system for fully automated analysis of apical 4C and 2C echocardiographic images can provide rapid and highly reproducible assessment of LVEF and LS.
This study utilized a recently developed fully automated offline analysis system for echocardiograms to assess reproducibility of LVEF and LS. Similar systems and algorithms have previously been evaluated and have demonstrated good intraobserver variability and correlation with reference methods typically with variability ranging from 1.3% to 7.9%. This study demonstrated an astoundingly low variability of 0%, which far exceeds that previously published. If confirmed over a variety of ultrasound platforms, this technology would represent a significant advancement in determination of ventricular function with echocardiography. Of note, this study addressed only reproducibility of the system when evaluating the same image set on two sequential occasions. Not addressed was the ability of the system to provide reproducible information on LV function when the same patient was evaluated with two separate platforms on the same day, or the same platform on two separate days. Determining a temporal and intermachine stability of this automated algorithm would be essential to allow widespread clinical application.
< Back to Listings