Mindboggle: Automated brain labeling with multiple atlases

BMC Medical Imaging

Table 2 Regions whose label agreement improves with multiple atlases

The effect of multiple atlases on label agreement varies by manually labeled region. Here we compare label agreements obtained by Mindboggle for each region by different numbers of atlases randomly selected from either of the two subject groups (up to 19 atlases for each of the 20 subjects). Numerical entries denote the minimum number of atlases that result in significantly higher label agreements than for single atlas data, by region (p ≪ 0.0001 for all regions except inferior temporal gyms: p < 0.01 and right temporal pole: p < 0.0002). For example, an entry of "9" means that significantly higher results were obtained using 9, 10, 11,... to 19 atlases versus using one atlas). The significance test is the same as that applied to the whole brain label agreement data.

ISSN: 1471-2342