Figure 9

Comparison between label agreements obtained with different numbers of atlases. A one-way ANOVA was performed to test if the means are the same for the label agreements obtained by the different numbers of atlases. A multiple comparison test was then performed using Tukey's honestly significantly difference criterion to determine which pairs of means are significantly different. The graph displays the mean for each number of atlases with a 95% confidence interval around the mean, based on the Studentized range distribution. If intervals are disjoint, their means are considered significantly different. The label agreement obtained with a single atlas is in blue and any significantly different result is in red or green. Green results are significantly higher than gray results (using three atlases). Using at least four atlases resulted in significantly higher label agreements and lower type II errors than when using one atlas (p ≪ 10^-6), suggesting that Mindboggle should be used with at least four atlases to benefit from the multiple atlas approach.