[gate-users] updated proposal and patches

[Claude Comtat]

Dear Long,

thank you very much for your effort to define a joint GATE - STIR 
proposal. I'll do my best to help you in this job. Indeed, the Gate 
developer team is more than happy to incorporate new contributions from 
different users.

To begin with, a few words about data format/organization in PET. By 
nature, in "3D PET" (acquisition  without septa or with the septa 
retracted), the data are actually 4D and not 3D. The four dimensions are:

1) radial
2) azimuthal
3) axial
4) polar

A sinogram is defined by dimensions 1 and 2, and the size of the third 
dimension (number of axial bins) decreases as the polar angle increases. 
In addition, because of the typical curvature of the detectors, the 
radial sampling is not uniform.  Theses facts make that there is no 
consensus on how to store these 4D data (unlike interfile for 3D data 
like in SPECT, where the polar angle is null and the radial sampling is 
uniform). Each manufacturer has its own implementation, highly dependent 
on the geometry of the scanner. For these reasons, the ECAT7 output 
could not be implemented as a generic output format for CylindricalPET 
and a new system had to be defined in order to support that format. I do 
not know how Interfile supports 4D PET data and ring curvature ??? For 
ECAT7 data, the non uniformity of the radial sampling is typically 
corrected during the reconstruction of the image (this also prevents 
truly generic reconstruction tools).

The sinogram output is closely related to the ECAT7 format, except for 
the polar mashing (span) and can not be implemented as it is in the 
cylindricalPET system. The radial, azimuthal, axial and polar indexing 
of the sinogram bin is based on the crystal numbering (approximate and 
non exact rebinning,  like for real ecat systems). For cylindricalPET, I 
would rather suggest a different approach, where the radial, azimuthal, 
axial and polar indexing of the sinogram bin is based on the exact 
radial, azimuthal, axial and polar coordinates of the LOR (exact 
rebinning). This will make also the reconstruction easier. The exact 
radial and azimuthal coordinates of the LOR are already calculated in 
the root output. Once you know the exact four coordinates of the LOR, 
based on the closest neighbor, you can assign the corresponding sinogram 
bin (you can let the user choose the sampling). I suggest to start with 
a root macro rather than a new Gate class (can be done later, when 
everything is working properly). This root macro will convert the 
listmode root output into an Interfile  sinogram format. I can help you 
in this job.

Best regards,

Claude


null99 at mails.tsinghua.edu.cn wrote:

>Dear all,
>
>The JOINT GATE STIR proposal is updated, including typo fix, and adding roadmap.
>Most important of all, adding request for help on GATE side implement.
>
>Patches including:
>stir patch
> 1. newly fixed support for userdefined scanner for ifheader_for_ecat7
>
>gate patch
> 1. introduce macro: B_JOINT_GATESTIR
> 2. add support for Userdefined_Scanner
> 3. add Interfile support for GATESinoToECAT7
>
>TODOs:
> STIR:
>   no plans yet.
> GATE:
>   request help on how to adding Interfile support for cylinder system,
>suggestions and contribution is highly appreciated, esp. those from GATE
>developer.
>   
>Again, I hope all kinds of questions, comments, suggestions, contributions! Thanks
>in advance!
>
>Regards,
>
>Long Zhang
>
>
>
>------------------------------------------------------------------------
>
>_______________________________________________
>gate-users mailing list
>gate-users at lphe1pet1.epfl.ch
>http://lphe1pet1.epfl.ch/mailman/listinfo/gate-users
>