A new three-dimensional echocardiographic (3DE) automated software (HeartModel) is now available to quantify left heart chamber. The aims of this study were to assess the feasibility of this technique in children; and its correlation with manual 3DE and cardiac magnetic resonance (CMR) for measuring left ventricular (LV) and left atrium (LA) volumes and LV ejection fraction (LVEF).

Ninety-two children (5 to 17 years) were prospectively included in two separate protocols. In protocol, 1, 73 healthy children (8.8±3.0 years) underwent 2D and 3D transthoracic echocardiography (EPIQ 7C, X5-1, Philips Healthcare). LV end-diastolic volume (LVEDV), LV end-systolic volume (LVESV), LVEF and LA volume at ventricular end-systole (LAV) obtained with automated 3DE (Fig. 1) were compared with the manual 3DE measurements. In protocol 2, automated 3DE measurements from 19 children with cardiopathy (12.8±2.9 years) were compared with CMR values. Test-retest, intraobserver and interobserver variability and the mean analysis time per patient were also examined for 3DE measurements.

Automated 3DE was feasible in 77% of datasets and reduced significantly the mean time per patient required for indices analysis compared with manual 3DE (20±2 versus 125±24seconds, P<0.0001), even when contour adjustment was performed (29±10, p<0.0001). In protocol 1, there was excellent correlation for LVEDV, LVESV and LAV between automated and manual 3DE (r=0.89 to 0.99, all P<0.0001) but less for LVEF despite contour adjustment (r=0.56 to 0.57, all P<0.0001). Compared with manual 3DE, automated 3DE without contour edit overestimated LVEDV, LVEF and LAV with small biases and underestimated LVESV with wider bias. With contour adjustment, the biases and limits of agreement (LOA) were reduced (biases: 0.9mL for LVEDV, −1.2mL for LVESV, 2.2% for LVEF; relative biases: 1.3% for LVEDV, −4.5% for LVESV). In protocol 2, there were excellent correlation for LV volumes and moderate correlation for LAV between automated 3DE with contour edit and CMR (r=0.76 to 0.94, all p<0.0003) but the correlation for LVEF remained weak (r=0.46, p=0.05). Compared with CMR, automated 3DE with contour edit slightly underestimated LVEDV and LVESV (relative biases: −7.2 to −7.8%), underestimated LAV with larger bias (relative bias: −31.6%), and had a negligible bias for LVEF (1.0%). However, LOA were clinically acceptable only for LVEDV and LVEF. Test-retest, intraobserver and interobserver variability for automated 3DE measurements were low (<12%).

HeartModel is a promising software for fast assessment of left heart chamber volume and function. Its feasibility in children aged more than 5 years is good, with high reproducibility. The automated 3DE measurements of LV and LA volumes are comparable to manual 3DE, especially when contour adjustment of automated 3DE values is performed. Compared with CMR, LVEDV and LVEF measured by automated 3DE with contour edit seem interesting in clinical practice.