Feasibility, accuracy and prognostic value of fully automated speckle tracking analysis-derived left ventricular ejection fraction and global longitudinal strain

Abstract

Abstract Left ventricular ejection fraction (LVEF) has been widely used for LV systolic assessment. However, it has drawbacks including large observer variability and unsatisfactory detectability of subclinical dysfunction. LV global longitudinal strain (LVGLS) by speckle-tracking on two-dimensional echocardiography (2DE) has been reported to be superior to LVEF. However, it may be influenced by the different methodology (manual versus semi-automated analyses) and biased study cohorts. A fully automated analysis can alleviate drawbacks of the LVEF and allow fair comparison between LVEF and LVGLS. We sought to evaluate the feasibility, accuracy of LVEF, and LVGLS measurements by the novel fully automated 2DE software against manual analysis and cardiac MRI feature-tracking (CMR-FT). Additionally, we tested prognostic utility of LVEF and LVGLS. In consecutive 436 patients undergoing CMR and 2DE, the fully automated analysis had excellent feasibility (97%). The correlation in LVEF and LVGLS between the fully automated analysis and the other two techniques was high ( r  = 0.82–0.95). During a median 26-month follow-up, 65 patients experienced cardiac events. In 422 patients successfully analyzed by the fully automated software (63 patients with cardiac events), both LVEF and LVGLS by the fully automated analysis were associated with cardiac events and their prognostic utility was not inferior to manual analyses. Nested Cox proportional hazard models and net reclassification analyses revealed fully automated analysis-derived LVEF and LVGLS did not have an incremental prognostic value over each other. A novel, fully automated analytical software demonstrated excellent applicability to clinical practice by providing reliable LVEF and LVGLS with comparable prognostic utility.