Dear Leon,
Reading that you'd like to segment surgical instruments, I think that shape
models are not the best way to do this:
The main advantage of shape models is that they can capture the variability of
an organ, but your instrument is always the same. Thus, a simpler template
matching would be better suited. Another point is the surgical instruments are
often elongated structures, but our shape model algorithm has been designed to
work best with compact shapes (as most organs are). For very elongated
structures, I do not expect good results with our method.
Regarding the surrounding organs you would like to segment, are these captured
fully in the image? Since field of view is typically small for US, I expect
organs to be only party visible in your images. For this application, our shape
model segmentation will not work.
In general, creating a shape model is a relatively complex process. We have
code that helps us doing that, but it's rarely an automated process. The first
thing you need is to determine correspondences, a task for which I have
published code at the Insight Journal (seehttp://hdl.handle.net/1926/224 ). I
can send you an updated version of this code if you are interested, but please
note that this is only the first step for creating a model. In general, you
will also want an appearance model, and we have no open source code for that
part.
I am sorry I cannot give you a more positive answer, but I do no think
statistical shape models will solve your problem. For any further question,
feel free to ask.
Best wishes,
Tobias
-----Ursprüngliche Nachricht-----
Von: Ren, Hongliang [mailto:[email protected]]
Gesendet: Montag, 7. November 2011 17:07
An: Heimann, Tobias
Betreff: Please help on creating a shape model for segmentationon of surgical
instrument
Dear Dr. Heimann,
As recommended by Dr. Nolden, I am writing for the possibility of using one of
the MITK module you developed.
I am very interested in using MITK for my project and thank you very much for
your excellent developments.
The module I am particularly interested is the shape model based organ
segmentation. I found the instructions
herehttp://docs.mitk.org/nightly-qt4/org_segment.html. I am using it to segment
surgical instruments and surrounding organs in a 3D ultrasound volumetric
image, so there is no such a shape model for particular instruments in the
current MITK. I wonder how I can create a shape model myself, which can be used
with MITK ?
Any suggestions are highly appreciated, and all your efforts or helps will be
credited. I can, of course, contribute my code/development back to the library
after my work.
PS. I noticed that you provide a brief instruction here 'The basic algorithm is
versatile and can be applied on all kinds of segmentation problems where the
structure of interest is topologically like a sphere (and not like a torus
etc.). If you are interested in other organs than those offered by the current
version of the Segmentation view, please contact our research team.'
Looking forward to hearing from you.
Thank you very much for any suggestions, Leon Ren
===================
Research Fellow
Department of Cardiovascular Surgery
Harvard Medical School& Children's Hospital Boston
Email:[email protected]
MSN:[email protected]
Boston, MA 02115, USA
(1) 617-855-8188 (T)
---------- Forwarded message ----------
From: Marco Nolden<[email protected]>
Date: Tue, Nov 1, 2011 at 6:05 AM
Subject: Re: [mitk-users] Please help on creating a shape model for segmentation
Cc:"[email protected]" <[email protected]>
Dear Leon Ren,
the shape model segmentation is a MITK-based application that is not part of
the public distribution. I recommend you contact the principal author of the
method, Tobias Heimann,[email protected], for discussing possible
collaborations.
Best,
Marco
Am 01.11.2011 04:00, schrieb ice tolo:
Dear MITK Researchers and Developers,
I am very interested in using MITK for my project and thank you very
much for your excellent developments.
The module I am particularly interested is the shape model based organ
segmentation. I found the instructions here
http://docs.mitk.org/nightly-qt4/org_segment.html. I am using it to
segment surgical instruments and surrounding organs in a 3D image, so
there is no such a shape model in the current MITK. I wonder how I can
create a shape model myself, which can be used with MITK ?
Any suggestions are highly appreciated.
PS. I noticed that you provide a brief instruction here 'The basic
algorithm is versatile and can be applied on all kinds of segmentation
problems where the structure of interest is topologically like a
sphere (and not like a torus etc.). If you are interested in other
organs than those offered by the current version of the Segmentation
view, please contact our research team.'
Thank you again,
Leon Ren
------------------------------------------------------------------------------
RSA(R) Conference 2012
Save $700 by Nov 18
Register now
http://p.sf.net/sfu/rsa-sfdev2dev1
_______________________________________________
mitk-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mitk-users
