2018 Volume 43 Issue 5 Pages 564-572
Purpose: Left ventricular (LV) chamber quantification using automatic endocardial border tracking has been introduced into clinical practice and it is attracting wide attention as a useful method with high accuracy and reproducibility. However, it remains unclear whether there is any difference in measurements among devices of different vendors.
Methods: This study compared LV chamber volume and ejection fraction (EF) measurements between different ultrasound machines manufactured by two vendors (A and B), in 74 patients in sinus rhythm (58±12 years old, 67 men, EF=52±14%, EF≤50% in 29 patients). End-systolic and end-diastolic volume (ESV and EDV) and EF were measured first automatically and then after manually correcting the endocardial border detection, using the two machines. Single regression analysis and Bland-Altman analysis were used to assess measurement errors.
Results: Although automatic LV chamber measurements generally showed a significant linear correlation with corresponding manual measurements, there were different patterns in systematic errors and in the effect of the manual correction on the errors between the vendors. For vendor A, although EDV and ESV were underestimated proportionally with an increase in LV volume (bias: −11.7 ml for EDV and −9.1 ml for ESV), EF was sufficiently accurate (bias: 1.2%). The manual correction produced a more accurate volume (bias: 2.5 ml for EDV and 1.1 ml for ESV) without any effects on EF measurement (bias: 0.7%). On the other hand, for vendor B, EF was overestimated proportionally with a decrease in EF (bias: 7.4%) owing to the underestimation of ESV (bias: −13.1 ml); the errors persisted even after equivalent manual correction procedure (bias: 7.5%).
Conclusions: There was an inter-vendor difference in systematic errors of LV volume and EF measurements and in the effect of the manual correction on the errors.