Stubburn Midline Dura - 7th tissue prior?

[dbrennan44]

Good evening ANTs experts,

I am running into a consistent issue when performing antsCT, which is that midline dura tissue (between the L/R visual cortices) is consistently included in brain mask and taken up as GM. I have tried altering my study template (to not include this dura), but I cannot get subject-specific templates to not also include this midline dura. My question is - would a possible solution to this be instead of missing the dura in the brain mask (proven to be fruitless), include the dura as a 7th tissue probability? Thank you in advance

[stnava]

Can you provide a picture? On Mon, Apr 26, 2021 at 6:53 PM Daniel Brennan @.***> wrote:

Good evening ANTs experts,

I am running into a consistent issue when performing antsCT, which is that midline dura tissue (between the L/R visual cortices) is consistently included in brain mask and taken up as GM. I have tried altering my study template (to not include this dura), but I cannot get subject-specific templates to not also include this midline dura. My question is - would a possible solution to this be instead of missing the dura in the brain mask (proven to be fruitless), include the dura as a 7th tissue probability? Thank you in advance

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/ANTsX/ANTs/issues/1181, or unsubscribe https://github.com/notifications/unsubscribe-auth/AACPE7UQ6TAGRIT2JZSJUCDTKXVIHANCNFSM43T2QH5A .

--

brian

[cookpa]

It's hard to exclude this structure from the brain mask because part of the brain masking procedure in antsBrainExtraction.sh is to perform a morphological dilation on the mask defined by registration, and this often fills out the gap between the left and right occipital lobes.

A seventh class could work. I don't think you'd need to do anything except to store the prior in the same place as the others, with the same file name pattern.

[dbrennan44]

@cookpa that makes sense

here is the "warpedPrior" brain mask (in lighter gray), with the posterior brain mask underneath. As you say, it inflates the mask to fill that space despite the prior warp getting it right.

@stnava , attached is the segmentation on the same axial slice. as you can see, antsCT considers this included dura as GM

I will try to use a 7th prior. This way, we are less dependent on the brain mask. Wondering what will happen with a less-prominent dura however...

[stnava]

This type of problem is present in lots of Different parts of the brain.

Would be very worthwhile to arrive at a general solution. Set of brain extractions that properly modeled the Dura/other brain boundary confounds not just here but also In other problematic areas like entorhinal and inferior frontal cortex, Which has a related but different problem

We pretty much know exactly how to solve this but lack the data and priors.

On Mon, Apr 26, 2021 at 7:30 PM Daniel Brennan @.***> wrote:

@cookpa https://github.com/cookpa that makes sense [image: BrainMask] https://user-images.githubusercontent.com/39420777/116163180-487f4300-a6c5-11eb-8197-0109c4d1d064.png

here is the "warpedPrior" brain mask (in lighter gray), with the posterior brain mask underneath. As you say, it inflates the mask to fill that space despite the prior warp getting it right.

@stnava https://github.com/stnava , attached is the segmentation on the same axial slice. as you can see, antsCT considers this included dura as GM

[image: BrainMaskSeg] https://user-images.githubusercontent.com/39420777/116163353-98f6a080-a6c5-11eb-9c70-bdfff6402387.png

I will try to use a 7th prior. This way, we are less dependent on the brain mask. Wondering what will happen with a less-prominent dura however...

— You are receiving this because you were mentioned.

Reply to this email directly, view it on GitHub https://github.com/ANTsX/ANTs/issues/1181#issuecomment-827205894, or unsubscribe https://github.com/notifications/unsubscribe-auth/AACPE7TJ5AQ7KA3WOSMGE6DTKXZQHANCNFSM43T2QH5A .

--

brian

[dbrennan44]

decent results using a 7th tissue prior

edit: I had to subtract the 7th tissue prior from the GM tissue prior to achieve these results. the GM prior overwhelmed the 7th "dura" prior in the resulting image when they both existed in this space