[Neuroimaging] Analyzing the topology of ROIs and flood-filling in python (skimage?)

Thu Sep 3 22:42:35 CEST 2015

This is great - I think that ultimately the solution very well be to expand
the ROI, as you propose.

But I am not still not sure that I follow your reasoning: are you saying
that what it only seems like a hole, but there is actually no hole in the
resulting deformed ROI? Note that I am summing across that dimension in
displaying the mask in cell 27 of the notebook I shared. As I understand
it, the only way there could be a 0-valued voxel in the middle of the ROI
is if there is a topological hole in the ROI.

Which leads me back to my original question: if I have an object
represented as a binary mask in a 3D array and I wonder whether it's
topologically a torus or a sphere, how do I go about calculating that?
Furthermore - why does ndimage.fill_holes not seem to fill that hole?
(maybe there's no hole? Is that what you meant?).

Thanks again!

Ariel

On Thu, Sep 3, 2015 at 8:24 AM, Jesus-Omar Ocegueda-Gonzalez <
jomaroceguedag at gmail.com> wrote:

> Hi Ariel,
> I checked the transformation and everything seems to be correct. The hole
> is an interpolation artifact. The problem is that the ROI is too thin (only
> 2 voxels), if you visualize the warped voxels as points in 3D you can see
> that there is no actual "hole", but it is generated by removing two voxels
> from the boundary of two different warped slices (a valid result from a
> diffeomorphic map). To illustrate this, I dilated the mask along the x-axis
> by one voxel (now the thickness of the mask is 3 voxels) and warped the
> dilated mask with the same transform, after doing this, I get this ROI:
> [image: Inline image 1]
> Anyway, the discussion about the transforms being only **approximately**
> diffeomorphic is still valid but it is not a problem in this particular
> case.
>
>
> On Wed, Sep 2, 2015 at 11:50 PM, Ariel Rokem <arokem at gmail.com> wrote:
>
>> Hi Omar,
>>
>> The other ROIs are here:
>> https://github.com/jyeatman/AFQ/tree/master/templates. I will think
>> about the rest of your response tomorrow!
>>
>> Cheers,
>>
>> Ariel
>>
>> On Wed, Sep 2, 2015 at 7:39 PM, Jesus-Omar Ocegueda-Gonzalez <
>> jomaroceguedag at gmail.com> wrote:
>>
>>> Actually, Ariel, nearest neighbor interpolation is a very unstable
>>> operation. If you interpolate at x or x+epsilon you may get different
>>> results for a very small epsilon, and discarding one single voxel may lead
>>> to a rejection of a large number of streamlines (I'm thinking about the
>>> boundary of the ROI too!, not only the "hole" ). I think it would be a more
>>> precise selection if you warped the streamlines to the template and select
>>> them there (now I see that we need that extension to the diffeomorphic map
>>> asap!).
>>>
>>> On Wed, Sep 2, 2015 at 9:19 PM, Jesus-Omar Ocegueda-Gonzalez <
>>> jomaroceguedag at gmail.com> wrote:
>>>
>>>> Thanks Ariel, and don't worry, this is very related to the work I'm
>>>> doing now, so this is actually very useful. I almost reproduced your
>>>> experiment, by any chance can you share: LOCC_ni, ROCC_ni and
>>>> midsag_ni?
>>>>
>>>> On Wed, Sep 2, 2015 at 8:29 PM, Ariel Rokem <arokem at gmail.com> wrote:
>>>>
>>>>> Hi Omar,
>>>>>
>>>>> Excellent - thanks so much for taking a look! I know that you are very
>>>>> busy these days, and so your attention on this is highly appreciated! I
>>>>> will try experimenting more with this, with different input parameters, as
>>>>> you suggested.
>>>>>
>>>>> If you also want to take a look, since #680 and #681 were merged into
>>>>> dipy, you can now run:
>>>>>
>>>>>     import dipy.data as dpd
>>>>>     MNI_T2 = dpd.read_mni_template()
>>>>>
>>>>> To get the template data.
>>>>>
>>>>> Thanks again,
>>>>>
>>>>> Ariel
>>>>>
>>>>> On Wed, Sep 2, 2015 at 6:16 PM, Jesus-Omar Ocegueda-Gonzalez <
>>>>> jomaroceguedag at gmail.com> wrote:
>>>>>
>>>>>> Hello guys!,
>>>>>> I have been working on this issue for some days now (this is very
>>>>>> interesting Ariel!, thanks for sharing your findings). Satra is totally
>>>>>> right that **in theory** the transformations should preserve the topology.
>>>>>> Unfortunately, the transformations are only **approximately**
>>>>>> diffeomorphic. I am totally sure that this issue should be there in the
>>>>>> original version of ants too (dipy's implementation is the same algorithm),
>>>>>> although maybe the new version (antsRegistration) may have some
>>>>>> improvements that I'm not aware of.
>>>>>>
>>>>>> Having said that, you can make the transforms closer to diffeomorphic
>>>>>> by reducing the `step_length` parameter (in millimeters) from
>>>>>> `SymmetricDiffeomorphicRegistration`, which by default is 0.25 mm. You may
>>>>>> try something about 0.15 mm. The objective is to avoid making very
>>>>>> "aggressive" iterations, so another way to achieve this is by increasing
>>>>>> the smoothing parameter from the CCMetric, the parameter is `sigma_diff`,
>>>>>> which by default is 2.0, you may try something bout 3.0 (I would first try
>>>>>> reducing the step size, though).
>>>>>>
>>>>>> I would like to try some other ideas, by any chance can you share the
>>>>>> data (MNI_T2)?
>>>>>> Thank you very much!
>>>>>> -Omar.
>>>>>>
>>>>>>
>>>>>>
>>>>>> On Wed, Sep 2, 2015 at 7:28 PM, Satrajit Ghosh <satra at mit.edu> wrote:
>>>>>>
>>>>>>> hi ariel,
>>>>>>>
>>>>>>> can you do nearest neighbor interpolation in
>>>>>>> `mapping.inverse_transform`? if your original ROI doesn't have holes and
>>>>>>> you are doing a diffeomorphic mapping, your target shouldn't have holes
>>>>>>> either. for a comparison you could run antsRegister and
>>>>>>> antsApplyTransforms, with nearest neighbor interpolation.
>>>>>>>
>>>>>>> cheers,
>>>>>>>
>>>>>>> satra
>>>>>>>
>>>>>>> On Wed, Sep 2, 2015 at 8:13 PM, Ariel Rokem <arokem at gmail.com>
>>>>>>> wrote:
>>>>>>>
>>>>>>>> Hi everyone,
>>>>>>>>
>>>>>>>> Jason and I are working on a port of his AFQ system (
>>>>>>>> https://github.com/jyeatman/afq) into dipy. We've started
>>>>>>>> sketching out some notebooks on how that might work here:
>>>>>>>>
>>>>>>>> https://github.com/arokem/AFQ-notebooks
>>>>>>>>
>>>>>>>> The main thrust of this is in this one:
>>>>>>>>
>>>>>>>>
>>>>>>>> https://github.com/arokem/AFQ-notebooks/blob/master/AFQ-registration-callosum.ipynb
>>>>>>>>
>>>>>>>> The first step in this process is to take a standard ROI of some
>>>>>>>> part of the brain (say, corpus callosum, which is where we are starting)
>>>>>>>> and warp it into the subject's individual brain through a non-linear
>>>>>>>> registration between the individual brain and the template brain on which
>>>>>>>> the ROI was defined (in this case MNI152). Registration works phenomenally
>>>>>>>> (see cell 17), but because this is a non-linear registration, we find
>>>>>>>> ourselves with some holes in the ROI after the transformation (see cell 27
>>>>>>>> for a sum-intensity projects). We are trying to use
>>>>>>>> scipy.ndimage.binary_fill_holes to, well, fill these holes, but that
>>>>>>>> doesn't seem to be working for us (cell 35 still has that hole...).
>>>>>>>>
>>>>>>>> Any ideas about what might be going wrong? Are we using fill_holes
>>>>>>>> incorrectly? Any other tricks to do flood-filling in python? Should we be
>>>>>>>> using skimage?
>>>>>>>>
>>>>>>>> Thanks!
>>>>>>>>
>>>>>>>> Ariel
>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> Neuroimaging mailing list
>>>>>>>> Neuroimaging at python.org
>>>>>>>> https://mail.python.org/mailman/listinfo/neuroimaging
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> Neuroimaging mailing list
>>>>>>> Neuroimaging at python.org
>>>>>>> https://mail.python.org/mailman/listinfo/neuroimaging
>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>> --
>>>>>> "Cada quien es dueño de lo que calla y esclavo de lo que dice"
>>>>>> -Proverbio chino.
>>>>>> "We all are owners of what we keep silent and slaves of what we say"
>>>>>> -Chinese proverb.
>>>>>>
>>>>>> http://www.cimat.mx/~omar
>>>>>>
>>>>>> _______________________________________________
>>>>>> Neuroimaging mailing list
>>>>>> Neuroimaging at python.org
>>>>>> https://mail.python.org/mailman/listinfo/neuroimaging
>>>>>>
>>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> Neuroimaging mailing list
>>>>> Neuroimaging at python.org
>>>>> https://mail.python.org/mailman/listinfo/neuroimaging
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> "Cada quien es dueño de lo que calla y esclavo de lo que dice"
>>>> -Proverbio chino.
>>>> "We all are owners of what we keep silent and slaves of what we say"
>>>> -Chinese proverb.
>>>>
>>>> http://www.cimat.mx/~omar
>>>>
>>>
>>>
>>>
>>> --
>>> "Cada quien es dueño de lo que calla y esclavo de lo que dice"
>>> -Proverbio chino.
>>> "We all are owners of what we keep silent and slaves of what we say"
>>> -Chinese proverb.
>>>
>>> http://www.cimat.mx/~omar
>>>
>>> _______________________________________________
>>> Neuroimaging mailing list
>>> Neuroimaging at python.org
>>> https://mail.python.org/mailman/listinfo/neuroimaging
>>>
>>>
>>
>> _______________________________________________
>> Neuroimaging mailing list
>> Neuroimaging at python.org
>> https://mail.python.org/mailman/listinfo/neuroimaging
>>
>>
>
>
> --
> "Cada quien es dueño de lo que calla y esclavo de lo que dice"
> -Proverbio chino.
> "We all are owners of what we keep silent and slaves of what we say"
> -Chinese proverb.
>
> http://www.cimat.mx/~omar
>
> _______________________________________________
> Neuroimaging mailing list
> Neuroimaging at python.org
> https://mail.python.org/mailman/listinfo/neuroimaging
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/neuroimaging/attachments/20150903/b214a5b9/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 4623 bytes
Desc: not available
URL: <http://mail.python.org/pipermail/neuroimaging/attachments/20150903/b214a5b9/attachment-0001.png>