[ITK-users] Ignore background pixels for gaussian filter computations

[Albert Gubern Mérida]

Hi,

I want to compute voxel features using Gaussian filters such as
LaplacianRecursiveGaussianImageFilter in breast x-ray images. However, I
get very strong responses at the skin line due to the large intensity
difference between tissue and backgound. I do have the segmentation of the
background. Is there anyway to exclude/ignore the background voxels from
the computation using a mask or setting a specific intensity value?

Thanks!

Best,

Albert

[Matt McCormick]

Hi Albert,

The result could be masked. Here's an example of one approach:


https://itk.org/ITKExamples/src/Filtering/LabelMap/MaskOneImageGivenLabelMap/Documentation.html

HTH,
Matt

On Mon, May 1, 2017 at 8:59 AM, Albert Gubern Mérida <

Post by Albert Gubern Mérida
Hi,
I want to compute voxel features using Gaussian filters such as
LaplacianRecursiveGaussianImageFilter in breast x-ray images. However, I
get very strong responses at the skin line due to the large intensity
difference between tissue and backgound. I do have the segmentation of the
background. Is there anyway to exclude/ignore the background voxels from
the computation using a mask or setting a specific intensity value?
Thanks!
Best,
Albert
_____________________________________
Powered by www.kitware.com
Visit other Kitware open-source projects at
http://www.kitware.com/opensource/opensource.html
http://www.kitware.com/products/protraining.php
http://www.itk.org/Wiki/ITK_FAQ
http://public.kitware.com/mailman/listinfo/insight-users

[Albert Gubern Mérida]

Thanks Matt! I was not aware of this filter. However, instead of masking an
image after filtering, I would like that the filter itself ignores a
specific region (i.e. background) for any calculation. Masking is not
enough if I apply large sigmas. I am also interested in making the filter
faster by ignoring computations in specified voxels.

Best,

Albert

Post by Matt McCormick
Hi Albert,
https://itk.org/ITKExamples/src/Filtering/LabelMap/
MaskOneImageGivenLabelMap/Documentation.html
HTH,
Matt
On Mon, May 1, 2017 at 8:59 AM, Albert Gubern Mérida <

Post by Albert Gubern Mérida
Hi,
I want to compute voxel features using Gaussian filters such as
LaplacianRecursiveGaussianImageFilter in breast x-ray images. However, I
get very strong responses at the skin line due to the large intensity
difference between tissue and backgound. I do have the segmentation of the
background. Is there anyway to exclude/ignore the background voxels from
the computation using a mask or setting a specific intensity value?
Thanks!
Best,
Albert
_____________________________________
Powered by www.kitware.com
Visit other Kitware open-source projects at
http://www.kitware.com/opensource/opensource.html
http://www.kitware.com/products/protraining.php
http://www.itk.org/Wiki/ITK_FAQ
http://public.kitware.com/mailman/listinfo/insight-users

[Matt McCormick]

Another approach: mask before passing the data to the filter, too.

Unless the unmasked region is very sparse, skipping the masked region in
general will not speed up the computation -- it may even slow it down,
depending on the implementation. This is due to CPU memory cache pipelines
and the need to treat mask boundaries in a special way.

HTH,
Matt

On Mon, May 1, 2017 at 2:14 PM, Albert Gubern Mérida <

Post by Albert Gubern Mérida
Thanks Matt! I was not aware of this filter. However, instead of masking
an image after filtering, I would like that the filter itself ignores a
specific region (i.e. background) for any calculation. Masking is not
enough if I apply large sigmas. I am also interested in making the filter
faster by ignoring computations in specified voxels.
Best,
Albert

Post by Matt McCormick
Hi Albert,
https://itk.org/ITKExamples/src/Filtering/LabelMap/MaskOne
ImageGivenLabelMap/Documentation.html
HTH,
Matt
On Mon, May 1, 2017 at 8:59 AM, Albert Gubern Mérida <

Post by Albert Gubern Mérida
Hi,
I want to compute voxel features using Gaussian filters such as
LaplacianRecursiveGaussianImageFilter in breast x-ray images. However,
I get very strong responses at the skin line due to the large intensity
difference between tissue and backgound. I do have the segmentation of the
background. Is there anyway to exclude/ignore the background voxels from
the computation using a mask or setting a specific intensity value?
Thanks!
Best,
Albert
_____________________________________
Powered by www.kitware.com
Visit other Kitware open-source projects at
http://www.kitware.com/opensource/opensource.html
http://www.kitware.com/products/protraining.php
http://www.itk.org/Wiki/ITK_FAQ
http://public.kitware.com/mailman/listinfo/insight-users