Abstract
New speckle tracking echocardiography (STE) tools allow an automated detection of strain measurements, but in case of suboptimal image quality or electrocardiogram signal, both regions-of-interest (ROIs) tracking and cardiac cycle’s landmarks need to be edited. In this study, we aimed to assess the variability of STE measurement performed with Tomtec’s Autostrain LV/RV/LA® application across different operators’ expertise in patients with significant aortic valve stenosis, known to have often suboptimal acoustic windows.
Automated strain analysis was performed by two observers with different levels of expertise in STE (a student sonographer and a trained cardiologist) on scans from 30 consecutive patients with moderate to severe aortic stenosis. Interobserver variability was tested. Intra-observer variability was also assessed repeating measurements about one month after the first set. Manual editing of the automated ROIs or cardiac cycle landmarks tracing results was manually made from both users on the majority (>80%) of examinations; notably RV strain analysis required the least editing. At repeated measurement test (Figure), the average strain values were not found to be significantly different for measurements of apical-four-chamber-global-longitudinal-strain (A4C-GLS) (mean difference 0.96%, P = 0.08), left-atrium-longitudinal strain (LALS) (peak-atrial-longitudinal-strain, PALS, mean difference −0.89% P = 0.37; peak-atrial-contraction-strain, PACS, mean difference −0.76%, P = 0.13) and right-ventricle-free-wall-strain (RV-FWS) (mean difference 0.44%, P = 0.59). A stronger agreement between testers was observed for patients with markedly reduced left-ventricle-global-longitudinal-strain (LV-GLS) and was not affected by the quality of acoustic windows. Similar trend was evident for the left atrial and right ventricular parameters.
New tool AutoStrain LV/RV/LA® for automated strain analysis showed a good agreement between operators with different levels of expertise in the challenging setting of patients with aortic valve stenosis. The reproducibility was particularly accurate for A4C view and in patients with worse LV-GLS values, independently from image quality.
