[Rtk-users] projection normalization

[Simon Rit]

Hi,
What I suggested / you're doing is what is called "flat field correction"
in my experience. Rings typically appear due to noise in the flat/flood
field image. I would suggest to use much better statistics to acquire the
flood field image, e.g., by acquiring 10 or 100 and averaging them. This is
what is often done in commercial scanners.
For monochromatic x-ray, there is no beam hardening problem. For
polychromatic x-ray, the rescaling is not obvious (HU is one solution) and
there will still be beam hardening problems.
Cheers,
Simon

On Thu, Jun 30, 2022 at 6:32 PM Howard <lomahu at gmail.com> wrote:

> Hi Simon,
>
> Thanks very much for your help in answering my questions. I have further
> follow up comments/questions (see below) and hope to understand the issues
> I encountered better. I will also include the offline discussion with
> another CBCT reconstruction expert.
>
> Regards,
> Howard
>
> On Tue, Jun 28, 2022 at 5:41 PM Simon Rit <simon.rit at creatis.insa-lyon.fr>
> wrote:
>
>> Hi,
>> 1) I think I would divide the measured projection by the flood field, no
>> reason to limit it to one scalar. The rest seems correct.
>>
>
> Yes, what I did before using the maximum intensity to scale the measured
> projection was incorrect. However, there are ring artifacts on the
> reconstructed images when using the flood field to do pixel by pixel
> correction. I learned from the offline expert that to correct for the ring
> artifacts I will need to do "flat field correction" or other techniques.
> For the projection data acquired from GATE simulation, are there any
> available RTK tools to do this correction? Any other suggestions would also
> be appreciated.
>
> 2) yes! But with a spectrum, it's hard to anticipate the ref value due to
>> beam hardening.
>>
>
> For monochromatic X-ray, we know the linear attenuation for the known
> materials at the given energy, therefore I should be able to rescale the
> grayscale values on the reconstructed images. However, there will still be
> beam hardening problems?
>
>
>> 3) It's "normal" to have negative values due to scatter,  not beam
>> hardening. But one should try to correct for scatter.
>>
>
> I see. This might be a bit tricky then. We'll have to explore the
> correction methods.
>
> Simon
>>
>>
>>
>> On Tue, Jun 28, 2022, 05:50 Howard <lomahu at gmail.com> wrote:
>>
>>> Dear RTK Users,
>>>
>>> I am having some questions on the correct way of normalizing/calibrating
>>> the projection images. My projections were generated using GATE (geant4
>>> simulation for those who may not know what GATE is) with cone beam X-ray
>>> and flat panel detector. Here are the steps I followed.
>>>
>>> 1. generate projections (proj) with a full rotation (360 degree, say 1
>>> projection/degree) with the imaging object in the X-ray beam path
>>> 2. remove the imaging object to generate an air projection (flood field)
>>> 3. Get the maximum intensity (maxInt_air) of the air projection obtained
>>> in step 2
>>> 4. Correct the projections obtained in step 1 with the formula proj_corr
>>> = -log(proj / maxInt_air) (I did see before some discussions using the
>>> maximum intensity to correct projection images.)
>>> 5. Reconstruct the CBCT using rtkfdk with the corrected projections
>>> obtained in step 4
>>>
>>> So here are my questions:
>>>
>>> 1)  Is the above procedure the correct way of reconstructing CBCT?
>>> 2) Are the grayscale values in the reconstructed CBCT correspond to the
>>> attenuation coefficients of those materials?
>>> 3) There are some negative grayscale values in the reconstructed CBCT
>>> due to artifacts such as the beam hardening. Is this normal?
>>>
>>> Many thanks for any feedback or suggestions.
>>>
>>> Howard
>>>
>>>
>>>
>>> _______________________________________________
>>> Rtk-users mailing list
>>> Rtk-users at public.kitware.com
>>> https://public.kitware.com/cgi-bin/mailman/listinfo/rtk-users
>>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/rtk-users/attachments/20220704/9ae491aa/attachment.htm>