Labeling errors. Fig. 6A demonstrates the variability in the spatial distribution of labeling errors for a single atlas labeling a subject, across all atlases. Blue indicates voxels where at least one atlas disagrees with the subject's manual labels (union). Green indicates voxels where every atlas disagrees with the subject's manual labels (intersection). Fig. 6B demonstrates the effect of the use of multiple atlases on labeling errors. Red voxels are those whose manually assigned label disagrees with the majority of the labels assigned by Mindboggle using multiple atlases. If we look from left to right, we see that increasing the number of atlases reduces labeling errors. Atlas selection and isosurface representation match the conditions of Figure 2.