Problem with processing T1w with noisy background

[philippn]

Hi all,

I am facing a challenge with processing T1w images that have a noisy background making it difficult to process them with 3dSkullStrip and @SSwarper (see image attached). Do you have any idea how to tackle this problem?

Happy to provide an example nifti.

Thanks and bests,
Philipp

[ptaylor]

Hi, Philipp-

Is that perhaps a quasi-synthetic T1w volume, made by combining multiple separate echos, and therefore the noise background is relatively elevated? If so, do you have the original component volumes?

–pt

[philippn]

Hi Paul,

thanks for your response! I think you are right, as there are also 3 other types of structural images that have been recorded (cf. file attached). Unfortunately, I don’t know how these images are being combined.

Bests,
Philipp

[storrisi]

sorry for chiming in

there’s quite a bit already out there on suppressing background noise in MP2RAGEs to make skull-stripping work, e.g.:
https://neurostars.org/t/mp2rage-background-removal-in-lemon-dataset/19987

-sam

[ptaylor]

Thanks, Sam, that is a useful link (and many of those replies do refer to @SSwarper directly).

The case of using a “component” dataset which should ~easily automask to punch away the background was what I was going to lean toward suggesting—the Case #4 I had listed in that other thread.

–pt

[philippn]

Thank you very much for the link, Sam and Paul for your comments!

I tried methods 4 and 5, as I have an image without noise for masking and the normalization after @SSwarper looks fine!

For some reason, however, does the skullstripped anatSS image not match the original anat anymore? See attached image.

Do you have an idea of what happened?

Bests,
Philipp

[ptaylor]

Was the original oblique?

3dinfo -obliquity DSET

?

[philippn]

That did the trick (see attached)!

Many thanks for both of you for your help!

Bests,
Philipp