A different way to segment MRI brain scan to WM, CSF and GM


Hello everyone!

For segmenting the MRI scan I can use FSL FAST. But, as it takes a lot of time to segment, I have thought of another way. So I need your suggestions on this.

I have the structural scan in subject space. I transformed it to standard MNI space using FSL FLIRT and got the transformation matrix (T). I can then take the standard CSF, WM or GM tissue prior and then transform it to subject space using T inverse. This will result in the tissue priors transformed to subject space which I can use to segment the structural brain scan. Will this work?

Thanks in advance!


Hi Varun,

That is not a bad idea for a rough quick approximation, but a structural scanner linearly registered to the MNI atlas can (and will) be wildly different in shape from the MNI atlas, so the priors will not work specially well (they will be ROUGHLY aligned with the structural image, but there will be plenty of differences).

The results would be better if you used non-linear registration (although not as good as using FAST), but in return, it will take way longer.

Therefore, I think FAST (or any other specific segmentation tool) is still worth using, rather than the FLIRT + priors idea.



Thanks a lot Fidel. I will try FAST and check how off the results are off with the FLIRT + Priors method.