Fmriprep: Input lesion mask to ANTs?

[wanderine]

We are trying to use fmriprep for brain tumour data and the registration seems to distort the brain tumours. Is it possible to include a lesion mask to be used by ANTs?

[DavidAbramian]

+1

[effigies]

Yes, lesion masking has been in the pipeline for a bit (poldracklab/niworkflows#121 + #359), but it hasn't been merged, yet. The code base has evolved somewhat since it was last worked on, so it's going to take a bit of work to get it to fit nicely with more recent changes. Developer time is spread a bit thin these days, but I think a one or two day sprint might be able to get this cleaned up and ready to go.

Possibly time can be made at Brainhack Global?

Related: #730

cc @danlurie

[danlurie]

A sprint during Brainhack Global would be great! I've been slowly trying to update things to match the new code in master, but it has been slow going. Having a team to work on it would be fantastic.

That said, this version should work: https://github.com/danlurie/fmriprep/tree/3c9283b70f23b26e5556d5443de6a128d1f2d086

[danlurie]

Is there interest in a sprint on this during BHG? If so, I can prepare some outlines/workflows/pseudo-code for how I've structured my code and what the final product should look like in terms of inputs/outputs.

[effigies]

Hi @danlurie, I'm definitely up for sprinting on this, but I should note that I'll only be at BHG-SF for Thursday night and Friday.

[danlurie]

Awesome! I'll be there Thursday night, most of Friday, and all of Saturday (I have to sneak out for an appointment around 11 on Friday). Something to consider is whether it is more efficient to try and update my branch or instead just re-implement the logic in the current code. I'll write up some stuff and we can chat more on Thursday night.

[effigies]

Hi @wanderine. We've merged lesion masking, and it will appear in the next release. I believe (@danlurie, can you verify?) that you can add a mask of the form: anat/sub-<participant_label>_roi.nii.gz to the directories of your participants with lesions, where the file is a binary mask with ones indicating the presence of a lesion and zeroes everywhere else.

We've discussed some changes to the locations to properly respect BIDS dataset boundaries (lesion masks are considered a derivative), so the exact handling will change as we finalize this approach (#1108).

In addition, we plan to add lesion masking to brain extraction, as well (#1109).

[danlurie]

Documentation for this feature can be found here until it is merged into the main branch.

[wanderine]

Thank you! We will try it out.

2018-05-05 16:32 GMT+02:00 Chris Markiewicz [e-mail]:

Hi @wanderine https://github.com/wanderine. We've merged lesion
masking, and it will appear in the next release. I believe (@danlurie
https://github.com/danlurie, can you verify?) that you can add a mask
of the form: anat/sub-_roi.nii.gz to the directories
of your participants with lesions, where the file is a binary mask with
ones indicating the presence of a lesion and zeroes everywhere else.

We've discussed some changes to the locations to properly respect BIDS
dataset boundaries (lesion masks are considered a derivative), so the exact
handling will change as we finalize this approach (#1108
https://github.com/poldracklab/fmriprep/issues/1108).

In addition, we plan to add lesion masking to brain extraction, as well (

1109 https://github.com/poldracklab/fmriprep/issues/1109).

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/poldracklab/fmriprep/issues/1056#issuecomment-386809710,
or mute the thread
https://github.com/notifications/unsubscribe-auth/AEGryOhtWDzF9q1Xx6gWH4Yu2LRF3q-Uks5tvbflgaJpZM4TTBX3
.

--
Anders Eklund, PhD