Comparison of automated versus semi-automated methods
Five hundred and thirty-eight datasets were used for automated and
semi-automated methods comparison from all centers. Bland-Altman
analysis showed wide limit of agreement (LOA) between the automated
versus semi-automated contouring of the 3D LV datasets. The bias (LOA)
for EDV, ESV, SV, and EF were -0.56% (-32.25 to 31.13), 1.01 (-29.94 to
31.96), -1.73 (-18.12 to 14.65), and -1.52 (-17.05 to 14.01)
respectively (Table 2). Sub-analysis of the inter-center reproducibility
for LV volumes and function between the core lab and other centers using
the automated method showed a bigger bias and LOA than the
semi-automated methods (Table 2). The inter-center reproducibility for
LV volumes and EF showed less bias and LOA between the core lab and
other centers using semi-automated contouring (Figure 1 and Table 2).
The percent difference using the automated and semi-automated methods
for LV EDV, ESV, and SV were summarized in Figure 2. The mean percentage
difference in EDV and ESV are higher using the automated method compared
to the semi-automated method. The mean absolute percent difference in 3D
LV EF between all centers was 7.4±5.4%. Figure 3 shows the Bland-Altman
analysis percent difference of the core lab and other centers using the
semi-automated method in ages 0-4 years, 5-9 years, 10-14 years, and
15-18 years of age respectively.