Semi-automated left ventricular endocardial detection versus hand-tracing in the measurement of left ventricular volumes and ejection fraction in daily clinical practice

Abstract

Background The common method of assessing left ventricle (LV) volumes and ejection fraction (EF) is hand-tracing Biplane Simpson method. Alternatively, ultrasound vendors offer different semi-automated LV endocardial border detection software with anatomical intelligence to assess LV volumes and EF. By using speckle-tracking technique, this software tracks the LV endocardium throughout the cardiac cycle and computes the LV volumes in every image frame using the disk summation method from which a volume-curve is generated, and the EF is calculated using the maximum and minimum volumes obtained. Data on the performance of this method in comparison with the hand-tracing Biplane Simpson method in daily clinical practice is scarce. Purpose To determine the accuracy of LV volumes and EF using semi-automated LV endocardial detection tracing, and to compare the reproducibility of this method with the hand-tracing Biplane Simpson method, among operators with varying level of experience in echocardiography. Methods This was a single center retrospective observational study, conducted in year 2020. 127 patients, aged >18 years, who underwent clinically indicated transthoracic echocardiography were recruited. The echocardiographic images were analyzed independently in a blinded fashion by 3 operators – a sonographer, a fellow-in-training and a cardiologist specialized in echocardiography. The LV volumes and EF were first measured using hand-tracing Biplane Simpson method, then repeated using semi-automated tracing at a different time and the operator were blinded to the initial hand-tracing measurements. Results The mean age of patients was 50±16 years, 35.4% were male, mean body surface area was 1.62±0.18m2, 92.1% were in sinus rhythm, and 61.4% had good acoustic window. Table 1 shows the LV end-diastolic volume (EDV), end-systolic volume (ESV) and EF, measured using different method, by the 3 operators. There were excellent correlation and agreement between semi-automated tracing measurements and hand-tracing measurements of LV EDV (r=0.985, LOA [mean ± 1.96 SD] 16.9 ml, ICC 0.991), ESV (r=0.990, LOA 12.7 ml, ICC 0.994) and EF (r=0.962, LOA 7.43%, ICC 0.967) by experienced cardiologist. The limit of agreement (LOA) between cardiologist and sonographer for semi-automated tracing measurement of LV EDV, ESV and EF were 29.13 ml, 19.74 ml and 9.25% respectively, which was comparable with that of hand-tracing measurement. The agreement between cardiologist and fellow-in-training for semi-automated tracing measurement of LV volumes and EF was slightly better than hand-tracing method, with a LOA of 25.60 ml, 17.48 ml and 7.08%, for EDV, ESV and EF respectively (Table 2). Conclusion In daily clinical practice, measurement of LV volumes and EF using semi-automated LV endocardial tracing method is accurate and demonstrates comparable reproducibility with hand-tracing Biplane Simpson method among operators with different level of experience in echocardiography. FUNDunding Acknowledgement Type of funding sources: None.