[Mristudio-users] LDDMM for Hematoma
Jeff Sadino
jsadino.queens at gmail.com
Thu Sep 8 23:02:30 EDT 2011
Hi Dr. Mori,
Thank you for your quick and detailed reply. It is good to know I am on the
right track by using the DWI image. I am confused by this line: "You can't
change the contrast ad hoc for LDDMM." Does this mean that I can only use
the B0, Trace, and FA images whenever I do the LDDMM registration? I
skullstripped both the template and subject DWI images. My thinking was to
use the skullstripped DWI images to drive the linear transformation, and
then (assuming a good AIR registration) use the linearly-registered subject
DWI image and the template DWI image to get a Kimap file that I could apply
to the rest of the linearly-registered subject images. I would do this for
this entire cohort of 17 subjects.
Or am I to read your response to mean that I should only do a linear
registration on the subjects and then stop there without doing an LDDMM
transformation? Or do I use the DWI image for the AIR, and then use the FA
and Trace for the LDDMM? Some of the subjects have hematomas in about
10-15% of the brain, so it is a significant region.
Thank you very much!
Jeff
On Thu, Sep 8, 2011 at 3:33 PM, susumu mori <susumu at mri.jhu.edu> wrote:
> Hi Jeff,
>
> Before you start LDDMM, you have to make sure the linear registration (AIR)
> is very good. This is similar to non-liner fitting of a XY plot, which
> always requires you to do linear fitting first to get a good initial
> estimate of the non-linear fitting parameters.
>
> There are cases when AIR fails often due to high intensity inside or
> outside the brains of patients. In this case, changing the contrast, as you
> did could be one solution. You can't change the contrast ad hoc for LDDMM
> but for linear transformation, it is just an initialization. So, all you
> want to do is to bring the two brains as close as possible. So, it's ok to
> use different approach for different data. What we often do is, 1) try
> mutual information, 2) use landmarks, or 3) do skull strip.
>
> Most important things is, to make sure that the linear transformation is
> good. Be careful about the angle of the brain and the brainstem. The
> sagittal rotation is one of the most difficult parameters, while X, Y, Z
> translation and coronal / axial rotation is usually easy to register. If the
> sagittal angle is not satisfactory, you should try different approaches.
>
> Also, we recently start to use 9-mode affine (no shearing) very often,
> which is now available in the latest DiffeoMap.
>
> Susumu
>
> On Thu, Sep 8, 2011 at 7:59 PM, Jeff Sadino <jsadino.queens at gmail.com>wrote:
>
>> Hello again DTI Experts,
>>
>> I have a group of patients that have suffered trauma to the head that I am
>> putting through the LDDMM pipeline. A problem I ran across during AIR in
>> DiffeoMap is that since the trauma is causing a hematoma in the B0 image,
>> the registration is very poor. On one subject, I performed DiffeoMap AIR
>> using the Mean DWI image and the registration came out pretty good. I am
>> wondering if this is a valid approach to perform AIR registration.
>>
>> For the LDDMM registration, usually I use the FA and the Trace image to
>> drive the Kimap. However, I think I will get the same poor results. Should
>> I use just the DWI image to drive the LDDMM registration? There would be no
>> need to do "Intensity Normalization" on the DWI image, correct?
>>
>> Thank you very much,
>> Jeff Sadino
>>
>> _______________________________________________
>> 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/20110908/2336e53c/attachment.html
More information about the mristudio-users
mailing list