[Insight-users] suggestions for ITK-based projects for a graduate image analysis class?

Thu Oct 25 10:41:57 EDT 2012

Alex,

In my experience, simplex meshes combined with deformable models are great tools for segmentation if they are dealt with properly.
The algorithm developed in my last segmentation project is a good proof!
http://www.biomedcentral.com/1471-2105/12/237

I truly welcome any additions to these classes which result in more efficient segmentation, so I will definitely be interested in being involved.
Having said that, I am gonna require more time as none of my current projects involve segmentation,

thanks to you and all the guys involved 

Leila

----- Original Message -----
From: Alexandre GOUAILLARD <agouaillard at gmail.com>
Sent: Thu, 10/25/2012 1:30am
To: Leila Baghdadi <baghdadi at phenogenomics.ca>
Cc: "Audette, Michel A." <maudette at odu.edu> ; insight-users at itk.org ; Wanlin Zhu <wanlinzhu at gmail.com> ; Bertrand MOREAU <bertrand.moreau at thecosmocompany.com> ; Humayun Irshad <humayun.irshad at gmail.com> ; Stéphane Rigaud <rigaud.stephane at gmail.com>
Subject: Re: [Insight-users] suggestions for ITK-based projects for a graduate image analysis class?

Dear michel, Leila, all

Leila code, and existing simplex mesh implementation is functional and
has been used for long by many in the community. It does the job. Now,
some expressed the need for additional functionalities.

What we are implementing right now is a dual mesh data structure for
oriented 2 manifolds. Most of the time people want to work with
surfaces of real-world (solid) objects, which are called by
mathematicians oriented 2-manifolds.
- However the original itk:mesh being n-dimensional, it does not
enforce consistent orientation nor 2-manifoldness. It is a problem
with deformable surfaces which tend to deform along the normals if the
normals are not consistently oriented. The solution for that has been
QEMesh.
- When working with simplex meshes the way it is implemented now, you
can work either with a triangular mesh, or with the simplex mesh.
Special filters have been developed for each, which do not work with
the others. The result is that often, you have a triangular mesh as
input, you use a filter to make the simplex mesh, you modify the
simplex mesh, your filter again to get teh triangular mesh, you apply
another filter, … If you trace the execution, you realize you spent
most of your time going back and forth the two representations. What
we wanted to do was to be able to modify either representation
(triangular or simplex) and have the other one automatically sync the
changes. You need two things for that: a data structure that can
handle both representation natively, and atomic (euler) operators that
can be applied to either representations.
- as a bonus, we wanted it to be as general as possible, so instead of
doing only triangular/simplexes, we aimed for primal/dual of which
triangular/simplex is a special case, but also voronoi/delaunay
- as a second bonus, we wanted to have not only subdivision (already
nicely implemented by wanlin, arnaud is right, we (read *I* ) should
take the time to bring it in the toolkit), but multiresolution, which
supposes to keep track of the successive subdivisions, somehow. Here,
benjamin gilles already has proposed a scheme for simplex meshes,
extending the original work of Delingette.

We are we right now. Thanks to great work by bertrand Moreau, Humayun
irshad and stephane rigaud, the data structure is implemented,
Primal/dual is implemented. Adaptor for writing and visualizing both
meshes from the data structure have been implemented. Low level
geometrical operators for exact Delaunay/voronoi have been
implemented.
Euler Operators and multiresolution have not been implemented yet,
pending hardening of the structure, test of all the existing mesh
filters, writing new tests, and writing the corresponding insight
journal papers. Michel's students in ODU are currently exploring using
the class in their projects.

Would you (leila, wanlin, arnaud, vincent's group at iowa) be ok with
spending a little bit of time playing with the beta quality code, as
you are maybe the most knowledgeable about surface meshes in ITK, and
as you might have great user cases?

Here are the already published IJ papers, and I attach the
not-yet-published ongoing version of the simplex mesh paper we should
publish early december. Feedbacks are welcome.
http://insight-journal.com/browse/publication/843
http://insight-journal.com/browse/publication/851
http://insight-journal.com/browse/publication/870

regards,

alex.

On Thu, Sep 27, 2012 at 12:11 AM, Leila Baghdadi
<baghdadi at phenogenomics.ca> wrote:
> Hello Michel,
>
> 1. My code from my paper, creates 32 single simplex meshes at the same time and and deform them individually (i.e, running 32 deformable model classes simultaneously) and uses binary masks (32 of them) created (also code I contributes to itk, TrignaleMeshToBinaryImage) to avoid collision between them, so I am not sure what you mean by "old v3 Simplex classes were not completely functional", all the functionality (create simplex sphere --> convert to and from Triangle to Simplex and deform Simplex mesh) that I used was there, I can not remember if I changed any of the simplex classes code but its easy to check the code.
>
> 2. sorry can not comment on Dr Gilles' work as I really do not know much about it,
>
> Leila
>
> ----- Original Message -----
> From: "Audette, Michel A." <maudette at odu.edu>
> Sent: Wed, 9/26/2012 11:53am
> To: Leila Baghdadi <baghdadi at phenogenomics.ca> ; insight-users at itk.org
> Subject: RE: [Insight-users] suggestions for ITK-based projects for a graduate image analysis class?
>
> Hi Leila,
>
> I'm familiar with Dr Delingette's work of course. I was referring to Dr Gilles' work on multi-surface Simplex, which builds on Delingette's single-surface, as well as provides new topological operators.
>
> My understanding about the old Simplex classes were that they did not completely implement the original Delingette model, which is why I'm working on a new Delingette Simplex implementation with the ITK Mesh class contributors (Alex Gouaillard and his colleagues), but this is separate from, and serves as a starting point for, my student's planned project. I may be wrong, or my understanding might not be up-to-date as it relates to v4, but my recollection based on my time at Kitware was that the old v3 Simplex classes were not completely functional; I'm surprised that you were able to build on them for your multi-surface code.  Please confirm.
>
> Best wishes,
>
> Michel
>
> Michel Audette, Ph.D.
> Assistant Professor,
> Department of Modeling, Simulation and Visualization Engineering,
> Old Dominion University,
> Norfolk, VA.
> ________________________________________
> From: Leila Baghdadi [baghdadi at phenogenomics.ca]
> Sent: Wednesday, September 26, 2012 11:40 AM
> To: Audette, Michel A.; insight-users at itk.org
> Subject: RE: [Insight-users] suggestions for ITK-based projects for a graduate image analysis class?
>
> regarding number 2, to my knowledge most of the simplex deformable code in ITK is related to the work of Herve Delingette (also from INRIA), so my code just uses that concept and elaborates on it for multi object segmentation,
>
> one other note regarding subdivision of mesh for segmentation of high curvature areas, the paper I published uses vtk linear subdivision method which I do not think its the right thing but later I heard about
> root 3 subdivision by Leif Kobbelt, I think its a great addition to the ITK deformable model segmentation which I hope to implement and submit it to ITK, so I will wait and see what changes you guys do to the original models,
>
> anyhow, I am all in for it, I can provide you with my code or help develop some of it, i.e the above
>
>
> Leila
>
> ----- Original Message -----
> From: "Audette, Michel A." <maudette at odu.edu>
> Sent: Wed, 9/26/2012 11:17am
> To: Leila Baghdadi <baghdadi at phenogenomics.ca> ; insight-users at itk.org
> Subject: RE: [Insight-users] suggestions for ITK-based projects for a graduate image analysis class?
>
> Hi Leila,
>
> I'm familiar with both topics.
>
> I became interested in FSL through my interest in FIRST, for deep brain stimulation applications.
>
> The kind of multi-surface simplex modeling that you describe is related to the work of Benjamin Gilles of INRIA, and Dr Gilles is helping my group with our applications of multi-surface simplex. I can tell you that I have an on-going collaboration that is looking at reimplementing the ITK Simplex implementation, and it may be of interest to use code that does multi-surface, in the manner suggested by Dr Gilles.
>
> If you have code that relates to either area  that is ITK-friendly, I think that we could use it, and it would be great to make a contribution via Insight Journal.
>
> Cheers,
>
> Michel
>
> Michel Audette, Ph.D.
> Assistant Professor,
> Department of Modeling, Simulation and Visualization Engineering,
> Old Dominion University,
> Norfolk, VA.
> ________________________________________
> From: Leila Baghdadi [baghdadi at phenogenomics.ca]
> Sent: Wednesday, September 26, 2012 10:58 AM
> To: Audette, Michel A.; insight-users at itk.org
> Subject: RE: [Insight-users] suggestions for ITK-based projects for a graduate image analysis class?
>
> Hi Michel,
>
> I was actually looking at a few of the ITK projects I have done in the past and wondering if I could find some free time to rewrite some of them for ITK4 and submit them to the journal,
> so not sure if this is exactly the right thing for your students but I am gonna give it a shot anyways,
>
> 1. BET (brain segmentation algorithm) http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/
> I wrote the code for ITK 3.2 sometime ago and found it very useful, infact I think it was used to segment preterm baby brains. not sure if this is in ITK yet
>
> 2. simplex deformable model segmentation (multi-object) --> I added extra code for multi-object and collision detection of multi-mouse embryo segmentation, also in ITK3.2
> published last year http://www.ncbi.nlm.nih.gov/pubmed/21679425
>
> Leila
>
> ----- Original Message -----
> From: "Audette, Michel A." <maudette at odu.edu>
> Sent: Wed, 9/26/2012 10:42am
> To: insight-users at itk.org
> Subject: [Insight-users] suggestions for ITK-based projects for a graduate image analysis class?
>
> Dear ITK users,
>
> I am teaching a graduate class in medical image analysis, and I would like to provide my students with a list of possible projects based on ITK that could implement recent algorithms, which possibly could then be submitted to Insight Journal and meet some needs of the ITK community. This class is based on the Toennies textbook from 2012, and will feature a number of registration and segmentation techniques, including graph cuts and level sets, as well as meshing methods not in the textbook. These projects will be for two students on which to collaborate, typically.
>
> If the community has any suggestions on how ITKv4 might be extended, based on recent papers in MICCAI, MedIA, or IEEE TMI, please let me know as soon as possible. Thanks for your kind consideration.
>
> Cheers,
>
> Michel
>
> Michel Audette, Ph.D.
> Assistant Professor,
> Department of Modeling, Simulation and Visualization Engineering,
> Old Dominion University,
> Norfolk, VA.
> _____________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Kitware offers ITK Training Courses, for more information visit:
> http://www.kitware.com/products/protraining.php
>
> Please keep messages on-topic and check the ITK FAQ at:
> http://www.itk.org/Wiki/ITK_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://www.itk.org/mailman/listinfo/insight-users
>
>
> --
>
>
>
> --
> BEGIN-ANTISPAM-VOTING-LINKS
> ------------------------------------------------------
>
> Teach CanIt if this mail (ID 705007640) is spam:
> Spam:        https://www.spamtrap.odu.edu/b.php?i=705007640&m=c69b8b1901f1&t=20120926&c=s
> Not spam:    https://www.spamtrap.odu.edu/b.php?i=705007640&m=c69b8b1901f1&t=20120926&c=n
> Forget vote: https://www.spamtrap.odu.edu/b.php?i=705007640&m=c69b8b1901f1&t=20120926&c=f
> ------------------------------------------------------
> END-ANTISPAM-VOTING-LINKS
>
> _____________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Kitware offers ITK Training Courses, for more information visit:
> http://www.kitware.com/products/protraining.php
>
> Please keep messages on-topic and check the ITK FAQ at:
> http://www.itk.org/Wiki/ITK_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://www.itk.org/mailman/listinfo/insight-users