[Insight-users] Hints to create a known deformation image - Deformable Image registration?

[motes motes]

Thanks for the detailed description! I know use the resulting
deformation field from the registration process on a regular grid to
get an impression of the deformation field. It works fine and gives a
good impression.

On Fri, Nov 6, 2009 at 10:34 PM, Luis Ibanez <luis.ibanez at kitware.com> wrote:
> On Tue, Nov 3, 2009 at 7:56 AM, motes motes <mort.motes at gmail.com> wrote:
>> On Mon, Nov 2, 2009 at 12:06 AM, Luis Ibanez <luis.ibanez at kitware.com> wrote:
>>> Hi Motes,
>>>
>>> 1) Lena is certainly not the best motif for evaluating a deformable
>>>     registration process. You may want to use a grid image with
>>>     about four time the density of lines of the BSpline grid, and
>>>     lines that are more than 2 pixels apart.
>>>
>>>     The reason why you want lines to be more than 2 pixels apart
>>>     is to avoid aliasing when you resample the image using the
>>>     deformation field.
>>
>>
>> I am still a bit confused about this suggestion. Do you suggestion to
>> use the grid image as input to the registration process?
>>
>> Why is Lena not a good example? I have made a deformable version of
>> the original Lena which I use as a moving image. I thought that the
>> more varying intensities an image contains the higher the chance of
>> alignment.
>>
> -------------------------------------------------------------------------------------
>
> Hi Motes,
>
> 1) Nope, I wasn't suggesting to use the grid image as input
>    to the registration process. On the contrary, I meant to
>    discourage you from doing so.
>
>    The grid image is good for illustrating the effect of the
>    resulting deformation field.
>
>    Typically what you want to do is to take two real images
>    register them, and the use the resulting deformation field
>    to resample the grid image. This will give you an idea of
>    the deformation that is applied to the moving image when
>    mapped in to the coordinate system of the fixed image.
>
>
> 2) Lenna is not a good example, "if" your real goal is to
>    perform registration of medical images.
>
>    The reason is that the spatial distribution of intensities in
>    the image of Lenna is quite different from what you see
>    in a medical image.
>
>    Lenna's image has large regions where the interaction of
>    the skin normal with lightning result in gradual variation
>    of intensity.  This type of intensity gradient is rare in medical
>    images of internal organs (except maybe when due to the
>    inhomogeneity of the field in an MRI image).
>
>
>  Regards,
>
>
>       Luis
>