Hi dear NITRC Community,

I am working on a review, where I compare different lesion delineation algorithms. Now I have collected and procecced data from different data sources. They are now in the same space (based on FLAIR): T1, and the lesion mask, also as brain-extracted, bias-corrected versions.

But now to my problem: I want to also compare some neural network tools, which need the same input space for training on all the different study protocols. 

Set 1 240 x 240 x 48, Resolution 1 x 0,958 x 0.958
Set 2 256 x 256 x 35, Resolution 1 x 1 x 1
Set 3 128 x 224 x 256, Resolution 1 x 1,25 x 1,03
Set 4 132 x 256 x 83, Resolution 1 x 1,2 x 0,97
Set 5 256 x 232 x 48, Resolution 1 x 1 x 1
Set 6 240 x 240 x 48, Resolution 1 x 0,958 x 0,958

I first thought about using FSL FLIRT with simply the MRI152 template, but I did not want to rotate the images and change with that the information per slice. I only want to change (resample, subsample) the coordinates to the same space and voxel size for all. 

I can also use nibabel, SPM or other tools. 

How would (or do) you approach this problem and is there an easy solution for that? For instance, has somebody tried out a whole-black image as reference using FLIRT in the target resolution? 

Best,
Niklas Wulms