Incorporating the "Active" (shape modelling) into Statismo

[maurici...@istb.unibe.ch]

Dear Statismo community,

Up to now Statismo has reached much enthusiasm and attention. We are very happy to see the interest and usefulness this library is providing.

 

There is now the interest to bring active segmentation into it by developing modules for shape adaptation. Regarding registration, Statismo can connect to the ITK registration framework through a dedicated class StatisticalShapeModelTransformation, see e.g. https://github.com/statismo/statismo/blob/master/Examples/ITK/shapeModelFitting.cpp , but it needs some more implementations regarding line search profile, statistical constraints, representation of intensity profiles, etc. 

It would be nice to know the interest from the community on this part and know who would be interested to join a working group to develop these functionalities. 

I hope my email raised your interest and look forward to hearing your comments and start a discussion on this point!

Best regards, 

  Mauricio

--

Mauricio Reyes, PhD
Group Head, Medical Image Analysis
Institute for Surgical Technology and Biomechanics
University of Bern
Stauffacherstrasse 78
CH-3014 Bern
Tel +41 31 631 59 50
http://www.istb.unibe.ch
http://www.mauricioreyes.me/
twitter.com/mreyesag
ch.linkedin.com/in/mauricioreyes/

[Rémi Blanc]

Hi Mauricio,

I see various applications as particularly interesting regarding model matching, for daily use of researchers in the field, but also for "end-users"
1/ shape model to (partial) target surface: this seems interesting for implant design/surgery planning/... even in the worst case where manual segmentation of the patient data is required. I think this one should be relatively easy to implement ; I have shareable source code for that, although it is not based on statismo and not so nice to read
2/ shape model to image, which I think is what you mean (really automated segmentation).
For this, I wonder to which extent elastix could help. I think there were movements in this direction ; Thomas?
Indeed, searching along the shape normals for a 'better' correspondence seems reachable when methods for extracting the line profiles, etc... are automated.
3/ Of course, in the end, having a composite model learned from shape & intensity, we could use this all for the matching, but maybe it's better to go step by step.

Now, should this be in statismo "core", or specific versions based on ITK /& VTK...

Cheers,

Rémi

--
You received this message because you are subscribed to the Google Groups "statismo" group.
To unsubscribe from this group and stop receiving emails from it, send an email to statismo-user...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Marcel Luethi]

Hi Remi and Mauricio,

The first use case (model fitting to a target surface) is already covered by statismo/ITK. One (of many) possibilities of how it can be done is illustrated in this example:
https://github.com/statismo/statismo/blob/master/Examples/ITK/landmarkConstrainedShapeModelFitting.cpp

The second use case, atlas based segmentation, is also already supported. It can either be done using ITK, or also by connecting statismo with elastix.

The elastic integration of statismo, which was done by Thomas, needs to be downloaded from a separate repository (https://github.com/statismo/statismo-elastix).
However, this solution currently require the use of a deformation model, as opposed to a shape model. Also, there is no support for using an intensity model in the fitting. In this respect, an Active Shape Model fitting  would be a big step forward.

I think the easiest way to implement it would be to use the ITK registration framework to implement the fitting algorithm, and add new data structures for the intensity models (e.g. profiles) to statismo. I would definitively support such an effort and help with the design, in case somebody is interested to start such a project.

Cheers,

Marcel

 

--
Marcel Lüthi, PhD
Department of Mathematics and Computer Science
University of Basel
Tel: +41 61 267 04 42

[Zhu Shanshan]

Hi Remi Blanc, 

I want to know do you have the code for searching along the shape normals for a 'better' correspondence ,i have already build a shpe model,and i want to use it to segment the target image,but i need the code for segmentation part,thanks!

在 2014年5月27日星期二 UTC+8下午8:43:56，Remi Blanc写道：

[Marcel Luethi]

Hi Zhu,

I have just published our Active Shape Modelling code to the repository

https://github.com/statismo/statismo-asm.

As mentioned before, the code is rather experimental and not yet polished. But all the main functionality should be there, and it should be relatively simple to adapt it to your own needs.

Best regards,

Marcel

On Thu, Dec 29, 2016 at 12:42 PM, 暖 <13005...@qq.com> wrote:

Hi,

 thanks very much,It is a great help for me. 

Waiting for your message always!

发自我的iPhone

------------------ Original ------------------

From: Marcel Luethi <marcel...@gmail.com>

Date: 周四,12月 29,2016 18:13

To: 暖 <13005...@qq.com>

Subject: Re: Incorporating the "Active" (shape modelling) into Statismo

Hi,

No worries about the copying. We plan to make it public anyway.

I will put the code in a separate github repository today or tomorrow and let you know once it is ready.

Best regards,

Marcel

On Wed, Dec 28, 2016 at 12:56 PM, 暖 <13005...@qq.com> wrote:

Hi Marcel,

Thank you very much for your reply, I am really in need of these codes, If you have time, I trouble you to send me these code for building and fitting Active Shape Models, I will  used these only for my own research,  not sharing with others.Thank you very much !

Finally, once again, thank you very much for your guide!

Zhu Shanshan

------------------ 原始邮件 ------------------

发件人: "Marcel Luethi";<marcel...@gmail.com>;

发送时间: 2016年12月28日(星期三) 晚上6:01

收件人: "暖"<13005...@qq.com>;

主题: Re: Incorporating the "Active" (shape modelling) into Statismo

HI Zhu Shanshan,

We have some code for building and fitting Active Shape Models that works with Statismo. We have not released it so far as it still mainly undocumented and the API is not yet finalized. If you are already familiar with C++ and Active Shape Modelling, I can provide it to you and you can take the bits that are useful for you. However, I will not be able to give you a lot of support, as I am currently rather busy.

in case you are flexible regarding the choice of programming language, you could also have a look at scalismo (github.com/unibas-gravis/scalismo). Scalismo supports ASM fitting out of the box and there is an online course (http://shapemodelling.cs.unibas.ch/StatisticalShapeModelling.html) which teaches the basics of shape modelling using scalismo (including how to fit Active Shape Models).

Best regards,

Marcel

To unsubscribe from this group and stop receiving emails from it, send an email to statismo-users+unsubscribe@googlegroups.com.

[zhaozh...@gmail.com]

Hi,Marcel! 

I want to ask a question about statismo-asm , Can I use the pcamodel.h5 generated by the command line statismo-build-shape-model  directly as input to fit-asm?  In other words, what is the modelname in fit-asm program's input?

在 2016年12月31日星期六 UTC+8下午7:33:39，Marcel Luethi写道：

[Marcel Luethi]

Hi Zhu,

An Active Shape Model is a point distribution model (which we call simply statistical shape model in Statismo) together with an intensity model.

The intensity model needs to be trained on a set of example data. You can find code to do that here:

https://github.com/statismo/statismo-asm/blob/master/src/itkBuildASM.cxx

The statismo-asm package has not been updated for quite a long time. I also don't know if anybody is actively using it at the moment. Therefore it is possible that it will not work out of the box and you will need to slightly adapt it to fit the latest statismo version.

Best regards,

Marcel