Table 10.
Results for model‐based, group‐wise and MAGeT Brain segmentations using atlases given as mean (confidence interval)
| Structure | Model‐based | Group‐wise | MAGeT Brain | |||
|---|---|---|---|---|---|---|
| ANIMAL | ANTs | ANIMAL | ANTs | ANIMAL | ANTs | |
| Kappa | ||||||
| Mouse | ||||||
| Hippocampus | 0.863 (0.850–0.881) | 0.847 (0.829–0.864) | 0.816 (0.797–0.836) | 0.833 (0.813–0.849) | 0.869## (0.853–0.884) | 0.850### (0.832–0.867) |
| Anterior commissure | 0.752 (0.732–0.772) | 0.751 (0.734–0.766) | 0.692 (0.672–0.710) | 0.728 (0.708–0.750) | 0.801***,### (0.790–0.811) | 0.759### (0.742–0.775) |
| Human | ||||||
| Striatum | 0.818 (0.808–0.827) | 0.816 (0.809–0.823) | 0.796 (0.787–0.804) | 0.807 (0.799–0.815) | 0.833**,### (0.825–0.837) | 0.845***,### (0.833–0.843) |
| Globus pallidus | 0.735 (0.712–0.752) | 0.740 (0.722–0.759) | 0.722 (0.697–0.746) | 0.728 (0.702–0.753) | 0.743 (0.723–0.762) | 0.752# (0.733–0.771) |
| Thalamus | 0.850 (0.838–0.856) | 0.848 (0.840–0.856) | 0.843 (0.833–0.854) | 0.849 (0.837–0.860) | 0.861# (0.881–0.900) | 0.856 (0.850–0.864) |
| Jaccard | ||||||
| Mouse | ||||||
| Hippocampus | 0.768 (0.740–0.789) | 0.735 (0.708–0.762) | 0.691 (0.671–0.710) | 0.712 (0.686–0.738) | 0.768## (0.745–0.792) | 0.740### (0.713–0.767) |
| Anterior commissure | 0.603 (0.578–0.629) | 0.601 (0.581–0.623) | 0.529 (0.507–0.550) | 0.574 (0.546–0.599) | 0.661***,### (0.653–0.683) | 0.611### (0.591–0.632) |
| Human | ||||||
| Striatum | 0.692 (0.683–0.702) | 0.690 (0.680–0.700) | 0.661 (0.649–0.673) | 0.740 (0.732–0.748) | 0.710**,### (0.702–0.719) | 0.740***,### (0.732–0.748) |
| Globus pallidus | 0.581 (0.557–0.606) | 0.591 (0.569–0.614) | 0.568 (0.538–0.597) | 0.610 (0.582–0.631) | 0.585 (0.570–0.620) | 0.610# (0.582–0.631) |
| Thalamus | 0.735 (0.723–0.749) | 0.738 (0.725–0.748) | 0.730 (0.714–0.745) | 0.731 (0.718–0.740) | 0.751# (0.740–0.766) | 0.741 (0.738–0.760) |
*,**, and *** represent significance levels of P < 0.05, P < 0.01, and P < 0.001 for MAGeT Brain as compared to its Model‐based counterpart.
#, ##, and, ### represent significance levels of P < 0.05, P < 0.01, and P < 0.001 for MAGeT Brain as compared to its group‐wise counterpart.
Only segmentations using the same nonlinear registration algorithms were compared to one another (e.g., only the model‐based segmentation using ANIMAL) was compared to the MAGeT Brain segmentation using ANIMAL.