<div dir="ltr">Hi again, <br><br>Thanks for your help Luis, I did the registration by using example DeformableRegistration2.cxx ( I could not find example Nr. 16 in any Release).<br><br>I changed the dimension to 3 and set the Nr. of iteration to 10 as Metric function does not improve considerebly if I use more iterations <br>
<br>Now my images look pretty steady in time which is a great improvement, however:<br><br>1)My volumes consists of 64 slices and out of 64 slices (zmin < z < zmax) my first slice (z=zmax) and last slice ( z=zmin ) (basically my top and my bottom slice ) are corrupted with some kind of noise. <br>
<br>Does that have anything to do with the registration itself? <br><br>2)Also I read that "Demons algorithm relies on the assumption that pixels representing the same homologous<br>point on an object have the same intensity on both the fixed and moving images to be registered"<br>
<br>If I understood this statement, this assumption does not hold in my case because intensity in my images is changing in time as contrast is flowing into the vessels, so same pixels (of the moving and fixed image) have good chances of not having the same intensity.<br>
<br>For this I was thinking of registering each volume (at time t=tn ) to its closest "registered" neighbour in time ( volume at time t=t(n-1)) in order to reduce the difference in the level of intensity. <br>
<br>Does anyone have a better suggestion ?<br><br>However this was already a good result,<br><br>Thank you for the help I will now consider to use VV.<br><br>Marco. <br><br> <br><br><br><br><div class="gmail_quote">2008/7/29 Luis Ibanez <span dir="ltr"><<a href="mailto:luis.ibanez@kitware.com" target="_blank">luis.ibanez@kitware.com</a>></span><br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">Hi Marco,<br>
<br>
Thanks for the clear description of your problem.<br>
<br>
1) You may find useful the VV visualization tool<br>
recently posted to the Insight Journal:<br>
<a href="http://insight-journal.org/midas/handle.php?handle=1926/1458" target="_blank">http://insight-journal.org/midas/handle.php?handle=1926/1458</a><br>
<br>
"VV: a viewer for the evaluation of 4D image registration"<br>
by Pierre Seroul, David Sarrut<br>
<br>
You could use it both for visualizing the 4D dataset<br>
as a 3D dataset animated in time.<br>
<br>
You could also use it to display deformation fields<br>
overlapping the 3D volume.<br>
<br>
<br>
2) Regarding the deformation itself. One method that<br>
you may want to try first is the multi-resolution<br>
Demons registration method.<br>
<br>
You will find an example on how to use this method<br>
in the file:<br>
<br>
Insight/Examples/Registration/<br>
DeformableRegistration16.cxx<br>
<br>
This example was kindly contributed by<br>
<br>
Vidya Rajagopalan,<br>
from the Bioimaging Systems Lab, at Virginia Tech<br>
<br>
during the NAMIC programming week on July 2008<br>
<a href="http://wiki.na-mic.org/Wiki/index.php/2008_Summer_Project_Week" target="_blank">http://wiki.na-mic.org/Wiki/index.php/2008_Summer_Project_Week</a><br>
<br>
<br>
NOTE: for your case,<br>
you must replace in line 204<br>
the Dimension value with 3 (currently is 2)<br>
<br>
<br>
<br>
Please note that image registration in general, and deformable<br>
registration in particular are ill-posed problems, and there<br>
is no perfect solution for them.<br>
<br>
It is more realistic to approach them as engineering challenges<br>
for which you need to identify a tool that does a satisfactory<br>
job under certain constratins of time and effort.<br>
<br>
Please give it a try at this example and let us know if you<br>
have further questions.<br>
<br>
<br>
Thanks<br>
<br>
<br>
Luis<br>
<br>
<br>
----------------------<br>
marco giordano wrote:<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"><div><div></div><div>
Dear all,<br>
<br>
I am a beginner with ITK and what I need to do is a registration of my 4D dataset which is corrupted by motion in time.<br>
<br>
In my images the patient is pushing his legs onto the scan-table during the acquisition and as a consequence I see the lower limbs as they were squashed as time goes on.<br>
<br>
My idea is to apply a deformable registration between the first volume (at time t0=0sec) and each one of the volumes at time t ( with t>t0)<br>
However as I do not have much experience with ITK I wanted to ask what is the most suited metric,optimizer, transform and so on to solve this registration problem.<br>
<br>
Thank you very much in advance <br>
<br>
-- <br>
Marco Giordano<br></div></div>
<mailto:<a href="mailto:MSN%253Agilmour812002@yahoo.it" target="_blank">MSN%3Agilmour812002@yahoo.it</a>><br>
<br>
<br>
------------------------------------------------------------------------<br>
<br>
_______________________________________________<br>
Insight-users mailing list<br>
<a href="mailto:Insight-users@itk.org" target="_blank">Insight-users@itk.org</a><br>
<a href="http://www.itk.org/mailman/listinfo/insight-users" target="_blank">http://www.itk.org/mailman/listinfo/insight-users</a><br>
</blockquote>
</blockquote></div><br><br clear="all"><br>-- <br>Marco Giordano<br><a href="mailto:MSN%3Agilmour812002@yahoo.it" target="_blank">MSN:gilmour812002@yahoo.it</a><br>ICQ :285-118-610<br>SKYPE:marcogiord81<br>
</div>