[Comparison of echocardiographic automatic border detection and magnetic resonance imaging. Measurements of the left ventricle in normal subjects].
Echocardiographic automatic border detection (ABD) has been the object of several studies with diverging results. The aim of this study was to verify the validity of ABD measurements by comparison with magnetic resonance imaging (MRI). Twenty healthy subjects underwent measurement of end systolic surface (ESA) and end diastolic surface areas (EDA) and fraction of surface variation (FSV), end systolic volume (ESV), end diastolic volume (EDV and ejection fraction (EF). These results were compared with the same parameters measured by cine MRI and a study of the variability of interpretation was performed on the echocardiographic parameters. An ABD analysis was possible in 80% of the study population. The correlations were satisfactory between the EDA and EDV (EDA; r = 0.84; SD = 1.9 cm2; EDV; r = 0.90; SD = 12 ml) with acceptable confidence intervals (CI) (EDA; [-4.02; 1.19 cm2]/EDV; [26; +7.9 ml]) and an underestimation of ABD values (EDA; -9%/EDV: -10%). With regards to the end systolic measurements, the correlations were not as good (ESA: r = 0.68; SD = 1.5 cm2/EDV: r = 0.59; +12 ml) with a more important measurement error (ESA: -2.05; +3.45 cm2)/EDA: (-9; +27 ml) and an overestimation of the ABD values (ESA; +10%; ESV; +18%). No correlation was observed between the FSV and EF. The intra and inter-observer errors were compared with those of conventional echocardiography (intra-observer error; 10.7-16.9%/inter-observer error; 10.8-16.6%). The authors conclude that ABD has a non-negligeable measurement error which limits its application in clinical practice. New transducers, automatisation of gain adjustment and new technologies should improve ABD measurements.