Artifacts in spiral cone-beam CT reconstruction

[zezisme]

@AnderBiguri Hello
I tried to reconstruct the spiral cone beam CT image using an iterative reconstruction algorithm (SART, OS-SART), as I understand it only the geo.offOrigin parameter needs to be modified to synchronize the position of the bed at each angle, and the other parameters are almost unchanged, but I get the following image, which has scimitar-like artifacts in the image;
In fact, I found that it should not be a problem occurring in the projection process (Ax), it should be a problem occurring in the inverse projection process (Atb), in addition I did the same reconstruction directly using the FDK algorithm, and the result is the same as OS-SART, with the same artifacts.
the geo i used


The spiral trajectory is as follows

I also used these same geometric parameters for the sheppLogan3D simulation FDK reconstruction and the results show that the reconstructed sheppLogan image has no artifacts
the projections(two angles) is

the FDK recon image(one slice) is (there is no artifacts)

But the same geo used in real data, got the atifact results.
here is the OS-SART algorithm result（real data）

And here is the FDK result

Do you have any idea what this could be due to?

Specifications

• MATLAB/python version: MATLAB
• OS:
• CUDA version:

[AnderBiguri]

It just looks like you have made a small mistake in the geoemtry, for example it could be that the direction of the angles should be the opposite, most likely (due to the moon shaped artifacts)

[zezisme]

It just looks like you have made a small mistake in the geoemtry, for example it could be that the direction of the angles should be the opposite, most likely (due to the moon shaped artifacts)

It doesn't seem to be due to the angles direction, I changed the angles direction just caused the reconstructed image to rotate 180°
add this line
angles = angles(end:-1:1);
here is the FDK result

Is it possible that the data is incomplete because only one loop was sampled (but why the sheppLogan3D simulation using the same geo can reconstruct the image without artifacts)

[AnderBiguri]

100% the issue is a geometric mismatch, those arcs only happen like that.
What about

Angles=-angles

[zezisme]

100% the issue is a geometric mismatch, those arcs only happen like that. What about

Angles=-angles

It still doesn't work.

I'll recheck the geometric as much as possible

[zezisme]

100% the issue is a geometric mismatch, those arcs only happen like that. What about

Angles=-angles

It is the mistake in the geo.offOrigin, i change it's direction, and it work!!!!
geo.offOrigin = -geo.offOrigin;

thank you very much!
(Surprisingly it is possible to reconstruct a spiral cone beam CT directly using TIGRE's FDK reconstruction algorithm, are the results reliable?)

[AnderBiguri]

Great! Glad it worked. For a small pitch fdk kinda works in helical, but I would not trust the quantitative values too much