[Rtk-users] Realistic Image Quality?

[Simon Rit]

Hi,
Attached is the python script using SimpleRTK
<http://wiki.openrtk.org/index.php?title=SimpleRTK> that I have used to do
the simulation of figure 6 of this publication
<http://www.creatis.insa-lyon.fr/~srit/biblio/rit2016.pdf>. The paragraph
where I describe what I'm doing starts with "The same simulations were
repeated with Poisson noise". The interesting part of the script for you is
line 158 where if I'm given a number of photons I0, I add Poisson noise.
Simon

On Wed, Jun 29, 2016 at 12:09 AM, Solomon Tang <solomoncztang at gmail.com>
wrote:

> Thanks for the feedback Chao and Simon,
>
> My geometry was using default RTK sdd/sid settings. I have now changed it
> to match the DICOM header from the original images (1085.6 SDD, 595 SID),
> but not much has qualitatively changed.
>
> How do you suggest adding photon noise? I have discovered an
> itkShotNoiseImageFilter but I'm not sure what is an acceptable scaling
> level. I'm assuming the reconstructed image should be passed through the
> filter, and not the projection.
>
>
>
> On Tue, Jun 28, 2016 at 1:34 AM, Simon Rit <simon.rit at creatis.insa-lyon.fr
> > wrote:
>
>> Hi,
>> I don't expect a drastic change, only a slight loss of spatial resolution
>> if the ray distance at the isocenter (I agree with Chao that it plays an
>> important role) is larger than the original voxel size. Maybe it's there
>> but you would need to zoom more to see it.
>> You would see a more realistic difference if you were adding photon noise
>> to your data.
>> Simon
>>
>> On Tue, Jun 28, 2016 at 9:14 AM, Chao Wu <wuchao04 at gmail.com> wrote:
>>
>>> What is the magnification factor of your geometry?
>>>
>>> 2016-06-27 23:56 GMT+02:00 Solomon Tang <solomoncztang at gmail.com>:
>>>
>>>> Hi Simon,
>>>>
>>>> I am using RTK to simulate CT acquisitions using different detector
>>>> sizes to see how this impact on image quality might change some of our
>>>> in-house metrics.
>>>>
>>>> The images I have linked to below have been created using
>>>> rtkforwardprojections with different projection spacings (0.3 isometric and
>>>> 0.75 isometric) reconstructed with rtkfdk with the same pixel spacing and
>>>> image dimensions (0.4688x0.4688x0.6 | 512x512x225). The CUDA projection
>>>> stepsize is equal to the projection spacing. The window levels between
>>>> images of their respective rows are the same.
>>>>
>>>> I am simply wondering if the differences between these images are
>>>> realistic. I would expect the image with a detector size than is more than
>>>> twice as large as the original would be drastically different when in fact
>>>> they turn out to be incredibly similar. Are the assumptions made about
>>>> projection spacing == cuda stepsize == simulated hardware detector size
>>>> incorrect?
>>>>
>>>> <http://goog_1486088111>
>>>> https://gyazo.com/e86436826f687a2db4b234699d050450
>>>>
>>>> https://gyazo.com/ca9612218f082e78ba3082950a27fa4c
>>>>
>>>> Solomon
>>>>
>>>> _______________________________________________
>>>> Rtk-users mailing list
>>>> Rtk-users at public.kitware.com
>>>> http://public.kitware.com/mailman/listinfo/rtk-users
>>>>
>>>>
>>>
>>> _______________________________________________
>>> Rtk-users mailing list
>>> Rtk-users at public.kitware.com
>>> http://public.kitware.com/mailman/listinfo/rtk-users
>>>
>>>
>>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/rtk-users/attachments/20160629/793b1682/attachment-0010.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: offsetRecon2.py
Type: text/x-python
Size: 6981 bytes
Desc: not available
URL: <http://public.kitware.com/pipermail/rtk-users/attachments/20160629/793b1682/attachment-0010.py>