A New Tool for Automatic Assessment of Segmental Wall Motion Based on Longitudinal 2D Strain

BACKGROUND: Identification and quantification of segmental left ventricular wall motion abnormalities on echocardiograms is of paramount clinical importance but is still performed by a subjective visual method. We constructed an automatic tool for assessment of wall motion based on longitudinal strain. METHODS AND RESULTS: Echocardiograms of 105 patients (3 apical views) were blindly analyzed by 12 experienced readers. Visual segmental scores (VSS) and peak systolic longitudinal strain were assigned to each of 18 segments per patient. Ranges of peak systolic longitudinal strain that best fit VSS (by receiver operating characteristic analysis) were used to generate automatic segmental scores (ASS). Comparisons of ASS and VSS were performed on 1952 analyzable segments. There was agreement of wall motion scores between both methods in 89.6% of normal, 39.5% of hypokinetic, and 69.4% of akinetic segments. Correlation between methods was r=0.63 (P<0.0001). Interobserver and intraobserver reliability using interclass correlation for scoring segmental wall motion into 3 scores by ASS was 0.82 and 0.83 and by VSS 0.70 and 0.69, respectively. Compared with VSS (majority rule), ASS had a sensitivity, specificity, and accuracy of 87%, 85%, and 86%, respectively. ASS and VSS had similar success rates for correct identification of wall motion abnormalities in territories supplied by culprit arteries. VSS had greater specificity and positive predictive values, whereas ASS had higher sensitivity and negative predictive values for identifying the culprit artery. CONCLUSIONS: Automatic quantification of wall motion on echocardiograms by this tool performs as well as visual analysis by experienced echocardiographers, with a greater reliability and similar agreement to angiographic findings.