I’ve simulated a uniform phantom in GATE and exported the GATE data to CASToR for image reconstruction. During reconstruction, I’ve chosen only the true events and ran a reconstruction, with and without attenuation correction. Then I visualize the image using ImageJ (see attached), but I do not understand why counts are appearing outside the phantom volume. Could anyone help me understand why this is happening here? Has anyone seen this before? Is it due to noise and/or scattering outside the phantom?
This is usually normal to have leftover signal outside the object, but it seems that you have more than normal.
How many iterations did you run ?
From your image, it seems that you are using a geometry without cylindrical symmetry.
If so, it may be possible that some regions of your field-of-view are not “seen” from a number of different angles that is enough to fully resolve these regions.
I think I’ve found the problem - it was due to the incorrect data format when visualizing the reconstructed image. I was using 2 iterations 20 subsets, MLEM, joseph projector, etc during the reconstruction. Please find the attached updated image, I think it is fine now. I’m now planning to perform other corrections, hopefully the scatter and random correction will clean up the reconstructed image and remove the dip in the center of the attached image.
Could you suggest me which method for scatter correction is relatively easy to implement? I’m thinking about the Single Scatter Simulation method or Triple Energy Window method for it. I would be grateful if you or someone guide me to do it.
Implementing the Single Scatter Simulation is no easy task. I never did it.
Depending on the final aim with your simulations, you may just ignore the scatters by not including them in the datafile.
About the attenuation corrected image, you should not have a dip but rather have a hill because of scatters that are not corrected. If there are only the trues inside the reconstruction, the cylinder should look flat except for high frequency variations due to the varying efficiency of the LORs.
The attenuation map is expected to be in cm-1, did you use mm-1 ?
