Left ventricular strain derived from computed tomography feature tracking: Determinants of failure and reproducibility

Abstract

ObjectivesTo evaluate the determinants of failure rate and reproducibility of computed tomography feature tracking (CT-FT) derived left ventricular (LV) strain. MethodsSubjects who underwent retrospectively electrocardiogram gated coronary computed tomography angiography (CCTA) were included in this study. CCTA image quality and strain curve were visually evaluated to determine whether the images were optimal for CT-FT strain analysis. Parameters of global strain, regional longitudinal strain (LS), territory LS, segmental LS, and LS mechanical dispersion were obtained. Reproducibility was quantified by intraclass correlation coefficient (ICC), Bland-Altman analysis, and coefficient of variation (CV). ResultsA total of 809 subjects were included in the final study cohort, and 625 subjects had complete LV CT-FT strain analysis. The risk factors for failed LV strain analysis were high average heart rate (odds ratios [OR] = 5.52; 95% confidence interval [95 %CI]: 3.59–8.68), high heart rate variation (OR = 2.7; 95% CI = 1.54–3.68), and older age (OR = 1.87; 95% CI = 1.24–2.84). Tube current modulation was inversely related to failure of LV strain analysis (OR = 0.19; 95% CI = 0.11–0.34). The intraobserver reproducibility of global strain was excellent (ICCs: 0.90–0.98, CVs: 3.4–9.7%), and the interobserver reproducibility was excellent (ICCs: 0.84–0.96, CVs: 4.4–11.9%), except for global radial strain. The intra- and interobserver ICCs of territory LS ranged from 0.83 to 0.97 and 0.76 to 0.95, while the CVs ranged from 3.7% to 9.2% and 4.6% to 10.6%.The intra- and inter-observer ICCs of regional LS ranged from 0.82 to 0.95 and 0.79 to 0.94, while the CVs ranged from 5.5% to 10.2% and 5.7% to 11.0%. The reproducibility of segmental LS was variable, and the ICCs were below 0.75 in 60.8% (31/51) segments. The reproducibility of LS mechanical dispersion was poor (the ICCs were below 0.50, and the CVs were greater than 20%). ConclusionsControlling heart rate during CT scanning is essential to perform a successful analysis of LV CT-FT strain. The reproducibility of CT-FT-derived global strain, regional LS, and territory LS was good to excellent, while the segmental LS should be used with caution in clinical practice, and LS mechanical dispersion is not recommended at present.