Add an option to output ids of voxels that are intercepted by the plane?

[wanqingy]

Hi!

I am actually looking for a way to get the cross sectional surface that is orthogonal to the direction of travel along the axons. More specifically, I need the outline of this cross sectional surface. I was so excited to see your package (through Forrest) but later realized that it's currently only output the area of the cross sectional surface.

Would it be possible to add an option to output all the voxel ids of the cross sectional surface?

Thanks!
Wan-Qing

[william-silversmith]

Hi Wan-Qing!

That's certainly something I can do. Would you prefer the output as a list of voxel locations or as an image where only the intercepted voxels are set to foreground?

The algorithm already does something similar to the latter internally, I sometimes extract it to visualize for debugging purposes.

Will

[wanqingy]

Imagery output would be great! Thanks, Will!

[william-silversmith]

Hi Wan-Qing,

The current master branch (not on PyPI yet) has a new function: xs3d.cross_section that returns a float32 image where each voxel's value is its contribution to the area. This should be easy to manipulate into whatever form you want. Let me know if that works for you and I can release it on PyPI.

Will

[wanqingy]

Yes, this works! Looking forward to try it out! Thank you! From: William Silversmith ***@***.***> Date: Friday, February 9, 2024 at 5:02 PM To: seung-lab/cross-section ***@***.***> Cc: Wan-Qing Yu ***@***.***>, Author ***@***.***> Subject: Re: [seung-lab/cross-section] Add an option to output ids of voxels that are intercepted by the plane? (Issue #5) Hi Wan-Qing, The current master branch (not on PyPI yet) has a new function: xs3d.cross_section that returns a float32 image where each voxel's value is its contribution to the area. This should be easy to manipulate into whatever form you want. Let me know if that works for you and I can release it on PyPI. Will — Reply to this email directly, view it on GitHub<#5 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/BCVKML73CSRM7XFXKFNP7TDYS3BK5AVCNFSM6AAAAABDBASS52VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSMZWG44DIOBTGE>. You are receiving this because you authored the thread.Message ID: ***@***.***>

[william-silversmith]

it's released! let me know if you run into any problems

…

On Mon, Feb 12, 2024, 3:00 AM Wan-Qing Yu ***@***.***> wrote: Yes, this works! Looking forward to try it out! Thank you! From: William Silversmith ***@***.***> Date: Friday, February 9, 2024 at 5:02 PM To: seung-lab/cross-section ***@***.***> Cc: Wan-Qing Yu ***@***.***>, Author ***@***.***> Subject: Re: [seung-lab/cross-section] Add an option to output ids of voxels that are intercepted by the plane? (Issue #5) Hi Wan-Qing, The current master branch (not on PyPI yet) has a new function: xs3d.cross_section that returns a float32 image where each voxel's value is its contribution to the area. This should be easy to manipulate into whatever form you want. Let me know if that works for you and I can release it on PyPI. Will — Reply to this email directly, view it on GitHub< #5 (comment)>, or unsubscribe< https://github.com/notifications/unsubscribe-auth/BCVKML73CSRM7XFXKFNP7TDYS3BK5AVCNFSM6AAAAABDBASS52VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSMZWG44DIOBTGE>. You are receiving this because you authored the thread.Message ID: ***@***.***> — Reply to this email directly, view it on GitHub <#5 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AATGQSKNAM2KQG2HNAPT2CDYTHDZ3AVCNFSM6AAAAABDBASS52VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSMZYGE4TKNZRGY> . You are receiving this because you commented.Message ID: ***@***.***>

[william-silversmith]

Hi @wanqingy! I released the feature we discussed in version 1.4.0 as projection = xs3d.slice(labels, position, normal)

So long as you don't care about the orientation of the projection, I think this should be pretty good for your purposes. If you need the orientation to be more similar to a standard basis let me know. I haven't figured out a good system for doing that yet (I'm using whatever pops out of a cross product rn).

[wanqingy]

Thank you, Will! I just tested it using binary image and it worked great! I will put into real work and keep you posted! From: William Silversmith ***@***.***> Date: Thursday, February 15, 2024 at 3:09 PM To: seung-lab/cross-section ***@***.***> Cc: Wan-Qing Yu ***@***.***>, Mention ***@***.***> Subject: Re: [seung-lab/cross-section] Add an option to output ids of voxels that are intercepted by the plane? (Issue #5) Hi @wanqingy<https://github.com/wanqingy>! I released the feature we discussed in version 1.4.0 as projection = xs3d.slice(labels, position, normal) So long as you don't care about the orientation of the projection, I think this should be pretty good for your purposes. If you need the orientation to be more similar to a standard basis let me know. I haven't figured out a good system for doing that yet (I'm using whatever pops out of a cross product rn). — Reply to this email directly, view it on GitHub<#5 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/BCVKML3FQQT2EQBOANJFMSTYT2IR3AVCNFSM6AAAAABDBASS52VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSNBXGQ4DMNRSGY>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[wanqingy]

Hi Will!

I tested it out for a real case and I think the projection image does not correct for anisotropic voxel resolution, right?

For a reference:
here is my code: https://gist.github.com/wanqingy/0d1a38cd852f42a49080e4bd9cee783e
here is my manual slicing: https://neuroglancer.neuvue.io/?json_url=https://global.daf-apis.com/nglstate/api/v1/4863895665639424

[william-silversmith]

You're right. The entire computation is performed in voxel space. For area computation, that's not such a big deal, you just change your normal to compensate I think, but for visualization... yea it becomes important again.

I'll have to look into this.

[wanqingy]

More important than visualization for me is that the width of myelin got squished when slicing off-axes. And it becomes inconsistent to detect myelin with my current strategy. I could probably find a way to compensate for it but if would be better if you have a way to account for anisotropy! Thanks as always!

[william-silversmith]

I think I may have fixed it!

What do you think? This is taken from minnie65. I believe the difference in orientation is due to the difference in how the vertical axis is displayed in Neuroglancer vs Microviewer.

Before anisotropy correction. (normal: [0,1,0])

After anisotropy correction. (normal: [0,1,0])

The location in minnie65:

https://neuroglancer-demo.appspot.com/#!%7B%22dimensions%22:%7B%22x%22:%5B8e-9%2C%22m%22%5D%2C%22y%22:%5B8e-9%2C%22m%22%5D%2C%22z%22:%5B4e-8%2C%22m%22%5D%7D%2C%22position%22:%5B122074.1875%2C96123.5%2C21339.3671875%5D%2C%22crossSectionOrientation%22:%5B0.5%2C0.5%2C-0.5%2C-0.5%5D%2C%22crossSectionScale%22:2.170592127183442%2C%22projectionScale%22:262144%2C%22layers%22:%5B%7B%22type%22:%22segmentation%22%2C%22source%22:%22precomputed://gs://iarpa_microns/minnie/minnie65/seg%22%2C%22tab%22:%22segments%22%2C%22segments%22:%5B%22864691135368664562%22%5D%2C%22colorSeed%22:3662937279%2C%22name%22:%22seg%22%7D%5D%2C%22selectedLayer%22:%7B%22visible%22:true%2C%22layer%22:%22seg%22%7D%2C%22layout%22:%22yz%22%7D

[wanqingy]

It looks great!!!

[william-silversmith]

Released in 1.5.0 lmk how it goes.

[wanqingy]

It worked for my data, but now it took me about 10x longer to get the projection image. Does it sound right to you?