Quantification of left ventricular dyssynchrony in patients with systolic dysfunction: A comparison of circumferential strain MR-tagging metrics

PURPOSE To define which circumferential strain MR-tagging metrics of left intraventricular dyssynchrony better identifies patients with systolic dysfunction against control subjects. MATERIALS AND METHODS One hundred fifty subjects were studied: (i) controls with ejection fraction (EF) > 55% (n = 84), (ii) patients with EF ≤ 55% not eligible for cardiac resynchronization therapy (CRT) (n = 52), and (iii) patients eligible for CRT according to the ESC guidelines (n = 14). Tagging cine MR-based circumferential filtered strain curves were extracted. Six dyssynchrony indices were studied: standard deviation (SD) of peak strain (SD_Ecc_ES), SD of time-to-peak (SD_TTP), strain delay index (LIM), regional variance vector (RVV), circumferential uniformity ratio estimate (CURE), and uniformity of strain TTP (US_TTP). RESULTS All metrics show significant differences between the three groups (ANOVA, P < 10(-4) ) and are correlated with EF. Significantly higher AUC values of ROC curves between patients with normal vs. decreased EF were obtained with SD_TTP (0.998) and CURE (0.995). Agreement among different methods was fair to good (kappa 0.32 to 0.89). Interobserver variability was best for CURE (1.2%) and US_TTP (0.8%) while more than 3-times larger for other metrics. CONCLUSION SD_TTP and CURE are the most discriminant dyssynchrony metrics for systolic dysfunction. However, taking into account the method's variability argues in favor of indices of uniformity of the strain, ie, CURE and US_TTP.