Is it fair to say that many of the ANTs tools were developed to work with images from any species? Or put slightly differently was ANTs developed to work with images from the human species?
On the contrary, it is my understanding that ANTs tools (e.g., antsRegistration and Atropos) are somewhat independent of imaging modality, imaged species (any type image for that matter) etc. I would just like to be more sure about this before responding to a reviewer who thinks ANTs was developed only for human work.
Thanks for your input,
Scott
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
ANTs was indeed developed as a general purpose tool intended for biomedical
imaging, in general. Valid across organ systems and species. You can point
to our work on the allen brain atlas, our contributions to the waxholm
template, our chimpanzee work (with Hopkins), our canine work ( with
Aguirre ) and all of Nick's work on the heart and lung. We also did well
(basically first overall) in a lung challenge that included cross-species
lungs and a segmentation challenge that included data from several
species/parts of the body.
Is it fair to say that many of the ANTs tools were developed to work with
images from any species? Or put slightly differently was ANTs developed to
work with images from the human species?
On the contrary, it is my understanding that ANTs tools (e.g.,
antsRegistration and Atropos) are somewhat independent of imaging modality,
imaged species (any type image for that matter) etc. I would just like to
be more sure about this before responding to a reviewer who thinks ANTs was
developed only for human work.
Hi Scott,
I use ANTs with rodent data on a regular basis and it works pretty well.
So far I have only found one instance of a human-centric assumption - in the antsMultivariateTemplateConstruction2 script, the parameters for N4 intensity normalisation use a spacing of several centimetres for the spline control points, which gives poor results for rodent brains due to their much smaller size. I circumvented this by multiplying up the image dimensions in the header.
Other than that, I have had significantly LESS problems using ANTs with rodent data than other pipelines (e.g. FSL or SPM).
Toby
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Nick,
I was concerned there might be another size assumption somewhere in antsMultivariateConstruction2 that I hadn't found (with hindsight I don't think there are). Other pipelines (notably SPM) make these assumptions all over the place so it's been standard practice at our site to multiply rodent dimensions up by a factor of 10 for years. Hence in my case it wasn't an issue to continue doing so.
Cheers.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi all,
Is it fair to say that many of the ANTs tools were developed to work with images from any species? Or put slightly differently was ANTs developed to work with images from the human species?
On the contrary, it is my understanding that ANTs tools (e.g., antsRegistration and Atropos) are somewhat independent of imaging modality, imaged species (any type image for that matter) etc. I would just like to be more sure about this before responding to a reviewer who thinks ANTs was developed only for human work.
Thanks for your input,
Scott
Hi Scott.
ANTs was indeed developed as a general purpose tool intended for biomedical
imaging, in general. Valid across organ systems and species. You can point
to our work on the allen brain atlas, our contributions to the waxholm
template, our chimpanzee work (with Hopkins), our canine work ( with
Aguirre ) and all of Nick's work on the heart and lung. We also did well
(basically first overall) in a lung challenge that included cross-species
lungs and a segmentation challenge that included data from several
species/parts of the body.
brian
On Mon, Jan 4, 2016 at 3:36 PM, Scott loves007@users.sf.net wrote:
Hi Scott,
I use ANTs with rodent data on a regular basis and it works pretty well.
So far I have only found one instance of a human-centric assumption - in the antsMultivariateTemplateConstruction2 script, the parameters for N4 intensity normalisation use a spacing of several centimetres for the spline control points, which gives poor results for rodent brains due to their much smaller size. I circumvented this by multiplying up the image dimensions in the header.
Other than that, I have had significantly LESS problems using ANTs with rodent data than other pipelines (e.g. FSL or SPM).
Toby
You don't have to change the image header to accommodate the smaller subject sizes. The -b option allows one to change the knot spacing.
Hi Nick,
I was concerned there might be another size assumption somewhere in antsMultivariateConstruction2 that I hadn't found (with hindsight I don't think there are). Other pipelines (notably SPM) make these assumptions all over the place so it's been standard practice at our site to multiply rodent dimensions up by a factor of 10 for years. Hence in my case it wasn't an issue to continue doing so.
Cheers.