I'm not quite sure what you are asking here as it isn't clear if you have a probabilistic atlas or a label atlas (the MNI atlas is probabilistic). It is possible to create a single 3D version of a 4D probabilistic atlas by choosing, at each voxel, the label with the maximum probability, often with an additional threshold, so that very low probability labels are excluded. This is what we store in our *-maxprob-thr??* images.

If this is what you want then there is a script to do it: $FSLDIR/data/atlases/bin/threshold

If, instead, you have a label atlas and want to to make a 3D version then you can just use fslmaths. If the individual ROIs already have unique numbers in the 4D version then just -Tmax will do what you want. If they do not have unique numbers then -Tmaxn -add 1 (and then mask it) will create a unique number based on which 4D volume the label appears in.

From: FSL - FMRIB's Software Library <[log in to unmask]> on behalf of Andrea <[log in to unmask]>
Reply-To: FSL - FMRIB's Software Library <[log in to unmask]>
Date: Thursday, 5 May 2016 21:51
To: "[log in to unmask]" <[log in to unmask]>
Subject: [FSL] build atlas with multiple ROIs

Dear experts in NeuroImaging,

I hope to find some recommendations using this helpful and vital forum.

I am trying to build a customized atlas using the atlases implemented in FSL (e.g. MNI atlas). Using the command

fslmerge -t <output> <input files>

this is working quite well.

However, I would love to build the atlas with several ROIs resulting in only one volume (4th dimension in the 4d file) instead of multiple merged files. In other software tools I recognized that this is possible. Examples are for example the aal90 atlas. All regions in the aal atlas are included in only one volume. This has some advantages I would like to benefit of as that using scripts all volumetric informations are in one single txt-output file instead of multiple files.

Any suggestions would be wonderful :)

Best, Andrea