ANTS Steps - Explanaing processing (not how-to)

These folders will be created (no need to create them):

(antsenv) root@REDWSONFAC26495:~# mkdir ~/rats_mri_project/
(antsenv) root@REDWSONFAC26495:~# cd rats_mri_project/
(antsenv) root@REDWSONFAC26495:~/rats_mri_project# mkdir template
(antsenv) root@REDWSONFAC26495:~/rats_mri_project# mkdir priors
(antsenv) root@REDWSONFAC26495:~/rats_mri_project# mkdir subjects
(antsenv) root@REDWSONFAC26495:~/rats_mri_project# mkdir outputs

Segmentation (Atropos)

priors as your initial, spatial expectations (“in this location GM is usually likely”), and the posteriors as updated, subject-specific probabilities after Atropos has looked at that person’s actual image intensities and applied the EM classification model.

     intensity model (mixture of Gaussians)
  p(I | class_k, θk)     ↑
                         │
prior_k(x)  ─────────► Bayes rule ──────►  posterior_k(x)
                + MRF smoothing

Priors (prior_k(x)) come from your atlas or numeric aliases
(rat_101_prior_01.nii.gz, …) and supply only spatial information, weighted here at 20 % (0.2 in the initializer).
The EM loop estimates intensity-class parameters θk from the subject’s image (-a $img_brain).
Bayes + MRF produce the voxel-wise posteriors:
P(class k | intensity I, location x, priors).

So yes—the posteriors are sharpened, subject-specific versions of the priors.