Fully Automated Evaluation of Left Ventricular Global Longitudinal Strain in Normal Hearts: An Assessment of Accuracy in the Pediatric Population

Two-dimensional echocardiography has been shown to be a useful modality for analyzing cardiac deformation parameters. However, the clinical utility of strain analysis is limited by time and training required for manual border tracing. This study aims to assess the feasibility of a "hands-off," measurement program, Philips AutoStrain, for measurements of left-ventricular deformation in children. We hypothesize that AutoStrain may accurately calculate deformation measurements, but its performance may be limited in certain patient populations. Echocardiographic images from 168 patients were analyzed via AutoStrain using fully-automated and semiautomated approaches. GLS were manually traced offline as a reference standard. Autostrain outputs were analyzed for accuracy using percent difference from the reference standard and an accuracy scoring system designed to objectively quantify the precision of Autostrain border tracings at the apex and mitral annuli. Both fully-automated and semiautomated AutoStrain correlated well with manual GLS (Fully-automated R = 0.71, Semiautomated R = 0.90). However, the accuracy of AutoStrain was affected by image axis, patient age, and clear visualization of the LV apex. Fully-automated techniques may lead to a "hands-off" future in assessment of LV deformation in older children, but should be used cautiously in younger children until further refinements in the algorithm are performed.