Hi,
It isn't easy to come up with an answer to such a question that always holds true. Some missing parts matter more than others and it is possible that FLIRT could be biased if some larger, consistent missing areas existed. Generally, the smaller the area the less it will matter to FLIRT, as it only has to determine the best translation and rotation for the whole brain, and so an artificial edge setup by the brain extraction being wrong would have to fight against all of the rest of the brain that would be aligned well. So although it can put things out of alignment, it usually doesn't do too much. The best way to know, as you suspected, is to run it and look carefully at the output.
One thing to be careful about though is if the brain extraction is being used for other processing (e.g., VBM, etc.) as then having missing portions of the brain can be a much worse problem.
All the best,
Mark
> On 9 Feb 2017, at 16:01, SUBSCRIBE FSL Jose <[log in to unmask]> wrote:
>
> Greetings,
>
> I have a question concerning Brain Extraction. Earlier (https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1701&L=FSL&F=&S=&X=1769E4BEDF7ADD8C06&Y=joseyordanramirez%40gmail.com&P=425440) I asked about what happens if you leave a bit of excess non-brain matter in the image, but now I'd like to ask what were to happen if I did the opposite and left some voxels out of the estimation. I have an example image in the attachment. The general recommendation is to NOT do that, but I'd like to know what happens if I do. Does the registration chop off the portion that isn't included? Does it warp the brain to fit the provided BET image? Would the boundary determination get messed up in other ways? I suppose this doubt stems more from not having a complete understanding about how the FLIRT algorithm works as a whole.
>
> A rule of thumb I've heard is that leaving out a small bit might be ok as long as it's only in one slice, but not if it's missing portions in several slices as in frontal region of the example image provided. (Actually the BET image is missing several portions now that I look at it again...) Are there any tolerable levels of 'missing' brain matter? I suppose the answer to this last question would be pragmatic (do it, then look and see), but that it makes more sense to use a parameter that doesn't miss any brain matter.
>
> Thank you very much,
>
> José
> <Missing_portions.png>
|