Int J Cardiovasc Imaging. 2016 Mar;32(3):429-37. doi: 10.1007/s10554-015-0804-x. Epub 2015 Nov 17.

Comparison of left ventricular manual versus automated derived longitudinal strain: implications for clinical practice and research.

Kobayashi Y^1,², Ariyama M^3,⁴, Kobayashi Y^3,⁴, Giraldeau G^3,⁴, Fleischman D^3,⁴, Kozelj M^3,⁴, Vrtovec B^3,⁴, Ashley E^3,⁴, Kuznetsova T^4,⁵, Schnittger I^3,⁴, Liang D^3,⁴, Haddad F^3,⁴.

Author information

1: Division of Cardiovascular Medicine, Stanford University School of Medicine, 300 Pasteur Drive Room H2170, Stanford, CA, 94305, USA. yukariko@stanford.edu.
2: Stanford Cardiovascular Institute, Stanford, CA, USA. yukariko@stanford.edu.
3: Division of Cardiovascular Medicine, Stanford University School of Medicine, 300 Pasteur Drive Room H2170, Stanford, CA, 94305, USA.
4: Stanford Cardiovascular Institute, Stanford, CA, USA.
5: Research Unit Hypertension and Cardiovascular Epidemiology, KU Leuven Department of Cardiovascular Sciences, University of Leuven, Louvain, Belgium.

Abstract

Systolic global longitudinal strain (GLS) is emerging as a useful metric of ventricular function in heart failure and usually assessed using post-processing software. The purpose of this study was to investigate whether longitudinal strain (LS) derived using manual-tracings of ventricular lengths (manual-LS) can be reliable and time efficient when compared to LS obtained by post-processing software (software-LS). Apical 4-chamber view images were retrospectively examined in 50 healthy controls, 100 patients with dilated cardiomyopathy (DCM), and 100 with hypertrophic cardiomyopathy (HCM). We measured endocardial and mid-wall manual-LS and software-LS, using peak of average regional curve [software-LS(a)] and global ventricular lengths [software-LS(l)] according to definition of Lagragian strain. We compared manual-LS and software-LS by using Bland-Altman plot and coefficient of variation (COV). In addition, test-retest was also performed for further assessment of variability in measurements. While manual-LS was obtained in all subjects, software-LS could be obtained in 238 subjects (95%). The time spent for obtaining manual-LS was significantly shorter than for the software-LS (94 ± 39 s vs. 141 ± 79 s, P < 0.001). Overall, manual-LS had an excellent correlation with both software-LS (a) (R(2) = 0.93, P < 0.001) and software-LS(l) (R(2) = 0.84, P < 0.001). The bias (95%CI) between endocardial manual-LS and software-LS(a) was 0.4% [-2.8, 3.6%] in absolute and 3.5% [-17.0, 24.0%] in relative difference while it was 0.4% [-2.5, 3.3%] and 3.4% [-16.2, 23.1%], respectively with software-LS(l). Mid-wall manual-LS and mid-wall software-LS(a) also had good agreement [a bias (95% CI) for absolute value of 0.1% [-2.1, 2.5%] in HCM, and 0.2% [-2.2, 2.6%] in controls]. The COV for manual and software derived LS were below 6%. Test-retest showed good variability for both methods (COVs were 5.8 and 4.7 for endocardial and mid-wall manual-LS, and 4.6 and 4.9 for endocardial and mid-wall software-LS(a), respectively. Manual-LS appears to be as reproducible as software-LS; this may be of value especially when global strain is the metric of interest.

KEYWORDS:

Dilated cardiomyopathy; Echocardiography; Global longitudinal strain; Heart failure; Hypertrophic cardiomyopathy; Post-processing software; Strain imaging; Vendor-independent; Ventricular function