Hello,

I am looking to create my own ROIs. Just verifying that they need to be created in the MNI space. Also unclear about the voxel size: is it 1X1X1 or 2X2X2?

Thanks,
Ana

Hi Ana,

There are two MNI space by default in FSL, but they are equivalent since the matrix size is halved in the case of 1mm in respect of 2mm. I attach below the details of this so that you can compare. You can use whatever you want if you are consistent later on with the transformations. But If you try to define ROI maybe it's better to use 1mm since it have 8 voxels for each 2mm voxel and you can be more accurate.

Dimensions: 91 x 109 x 91
Voxel size: 2 x 2 x 2

Dimensions: 182 x 218 x 182
Voxel size: 1 x 1 x 1

Hope to be helpful.

Best,

Magí.

Thanks so much, my fault I haven't mentioned I was using SPM, but I guess the same principles apply. Ana

I'm not used to SPM, but it depend on what "atlas" it uses. Anyway, just take into account that, for example in 2D, a 4x4 matrix of size 2mm can be a 8x8 matrix of 1mm, it's not the same, but equivalent because each pixel of 2mm will be mapped in 4 1mm pixel group. Just be coherent when apply transformations and all it will work.

Good luck!

Hi Ana,

You can technically work in any space that you want - although MNI is a good one because it is used across a lot of software, and you know, it's not Talairach :O)  I would say to choose the space that matches the analysis/software you want to use.

As for the voxel size, it almost doesn't matter as long as you have a standard anatomical of the same size and dimension that you could use to register to another voxel size, and then apply the registration to your ROI.  I could see arguments for wanting either resolution - if you would be flattening your images into a matrix the smaller file (larger pixel) would be ideal, but standard analysis on smaller numbers of images would be fine with the latter.  Again, maybe it just makes sense to choose the highest resolution, and then downsample if needed?

Lastly... the software! You can of course save an analysis contrast or group map as a mask, or use PickAtlas to create a more custom one based on shapes or anatomical regions.  One of my favorite ROI-making experiences was actually in MRIcron because I could click though every slice and I felt like I was crafting a work of art!  If you want to make tissue specific ROIs (for example, for different tissue types for glioblastoma multiforme) I like to use Mipav because you can click through layers, color in your tissues of interest, and it does a very good job to grow the region.

I hope that helps!

Best,

Vanessa

Many thanks, got an idea of what I need to do. Ana