[Gate-users] STIR-strange hot voxel in the center

[Kris Thielemans]

Hi Melanie

> I´m using STIR as reconstruction tool. In the reconstructed 
> images there is
> always a hot voxel right in the center of the FOV apparent, 
> which does not
> belong to the imaged phantom. This effect does not occur when using
> iterative reconstruction methods instead of FBP. I´m not sure 
> if this is a
> problem arising from STIR or from the measured data it self.
> Any ideas or suggestions would be appreciated.
> 

This might be more a topic for the STIR list, but anyway.

are you sure you don't have the known problem with the default backprojector
in STIR (incremental interpolating). This shows up depending on your
processor/compiler and compiler settings and usually gives rise to a bright
pixel in the middle, sometimes also as lines at 45 degrees.

Easiest way to find out is to run the STIR recon_test_pack.

If this is your problem, here a few (unsatisfactory) options:
- ignore the central voxels :-)
- replace the default backprojector with another one, as documented in STIR.
For FBP you can use the ray tracing matrix (But with a lot of samples per
LOR). For FBP3DRP this isn't entirely correct due to Jacobian factors. You
can use the interpolating matrix, but it's very slow (until it's cached).
You can solve that by writing the matrix to disk once, and then re-using
that saved matrix. That'll be pretty fast.
- change the optimisation flags of your compiler. That'll usually slow the
executable down though.

Do some searching on the stir mailing list for more info on this topic.

Kris