[Mristudio-users] DTI normalization with Diffeomap on lesioned brains

[susumu mori]

You need to mask anatomically equivalent regions in both the patients and
the atlas, which is difficult.
I assume your "lesions" mean surgically removed tissues. Is it correct? If
you can assume that the anatomical deformation around the lesions is small
(like the partial moon), you can register the patient image to the atlas by
affine (either automatically or landmark-driven) and transfer the lesion
locations to the atlas manually, I think. In stead of affine, you may be
able to use low-elasticity LDDMM (alpha = about 0.01).

I didn't quite get your second question. Using LDDMM (DiffeoMap), you can
control the elasticity by alpha value. 0.01 and larger give low elasticity.
0.005 is recommended for normal brains. 0.001-0.002 is used when anatomical
difference is large.

When you want alpha smaller than 0.005, cascading alphas are recommended
(e.g. three consecutive LDDMM with 0.01-0.005-0.002) to avoid being trapped
by a local minima. Using this cascading-alpha approach, you can also obtain
the intermediate results; results with low elasticity (alpha = 0.01),
medium alpha (alpha = 0.01/0.005).

In DiffeoMap, you can use up to 3 cascading alphas.

On Fri, Nov 30, 2012 at 7:54 PM, Dorian P. <alb.net at gmail.com> wrote:

> Thank you for answering Dr. Mori,
>
> Would it be feasible to mask only the lesion on the patient or does it
> have to be on both patient and atlas?
>
> Has anybody compared a T1 normalization on SPM and applying the matrix to
> DTI to a low elasticity normalization in DTIstudio?
>
> Best regards.
> Dorian
>
>
> 2012/11/30 susumu mori <susumu at mri.jhu.edu>
>
>> In general, lesions (severe intensity changes) or severe anatomical
>> changes (e.g. missing tissues or tumor) are very difficult to deal with,
>> when you are using a transformation-based analysis.
>>
>> When you use transformation with low elasticity (e.g. linear
>> transformation), the transformation could be less affected because the
>> registration would be mostly driven by the brain outlines. So, if your
>> lesions (intensity changes) do not accompany anatomical deformation, you
>> may be able to compare patients and controls.
>>
>> As the elasticity increases, the transformation would be more affected by
>> the lesions. Of course, if the lesions are small, they may not cause much
>> problems but I assume you have large lesions.
>>
>> If you are studying, for example, stroke and interested in anatomical
>> changes of contralateral hemisphere (or the brainstem), you can mask-out
>> the ipsilateral hemisphere for both patients and the atlas. We have done
>> this type of analyses before.
>>
>> Sorry that I could give you only a generic answer. This is an issue that
>> the whole community faces.
>>
>> On Fri, Nov 30, 2012 at 2:52 PM, Dorian P. <alb.net at gmail.com> wrote:
>>
>>> Hi all,
>>>
>>> I am interested on normalizing lesioned brains with a dual channel
>>> Diffeomap procedure. The normalization will be done on MNI existing
>>> template of Diffeomap. Lesions typically come from surgical resections.
>>>
>>> Does this procedure work well on lesioned brains or shall I need some
>>> mask for Diffeomap to avoid normalizing the lesioned area?
>>>
>>> Any comment is welcome. I couldn't find much about the topic on previous
>>> posts and articles.
>>>
>>> Thank you.
>>> Dorian P
>>>
>>> _______________________________________________
>>> mristudio-users mailing list
>>> mristudio-users at mristudio.org
>>> http://lists.mristudio.org/mailman/listinfo/
>>> Unsubscribe, send a blank email to:
>>> mristudio-users-unsubscribe at mristudio.org
>>>
>>>
>>
>> _______________________________________________
>> mristudio-users mailing list
>> mristudio-users at mristudio.org
>> http://lists.mristudio.org/mailman/listinfo/
>> Unsubscribe, send a blank email to:
>> mristudio-users-unsubscribe at mristudio.org
>>
>>
>
> _______________________________________________
> mristudio-users mailing list
> mristudio-users at mristudio.org
> http://lists.mristudio.org/mailman/listinfo/
> Unsubscribe, send a blank email to:
> mristudio-users-unsubscribe at mristudio.org
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.mristudio.org/pipermail/mristudio-users/attachments/20121201/68b8982a/attachment-0001.html