[Castor-users] GATE ecat reconsstruction

[STUTE Simon]

I was talking about the number of crystals in a single ring of crystals. From what I understand you have 62 x 5 crystals in a ring = 310. So do not use more than 16 subsets. Then try with 3 iterations first to see what you get. And use a post-smoothing like "-conv gaussian,3.,3.,3.::post" (FWHM of 3 x 3 mm).

________________________________________
De : Dimitrios Thanasas <dthanas at uoa.gr>
Envoyé : lundi 15 avril 2024 13:57
À : STUTE Simon
Cc : castor-users at lists.castor-project.org
Objet : Re: [Castor-users] GATE ecat reconsstruction

Dear Simon,

Thank you very much for your response.

I simulated a PET system with 62 blocks (xlength:4cm, ylength:5cm,
zlength:20cm around a ring of 50cm minimum diameter - 54cm maximum.
Each block has crystals of xlength:4cm, ylength:1cm, zlength:1 cm, so
there are 100 crystals in each block.
The source is orthogonal parallelepiped (xlength:5mm, ylength:1mm,
zlength:10mm).
I tried to  run (I do not know how to use post-smoothing):
castor-recon -df ecat_df.Cdh -opti MLEM -it 50:310 -dim 100.,100.,100.
-vox 1.,1.,1. -dout myimages1
but I get almost the same image as before.

310 is the total number of crystals (6200) divided by 20, 50 iterations I
think are too many.
As I want to see the thin side of the source (1mm), I think that I have to
use crystals with side 1mm and not 1cm as I am using now.

Best regards,
Dimitris


> Hi,
>
> Yes likely multiple problems.
>
> You should provide details about your detector in order to be able to
> understand the choice of some parameters.
>
> Why 64 itérations with 128 subsets ?! A rule of thumb without TOF would be
> to have a maximum number of subsets equal to the number of crystals in a
> ring divided by 16 to 20. Then, without any post-smoothing or
> regularization, aim for a maximum of 50 to 100 updates (i.e. iterations x
> subsets). For your first reconstructions, consider including a
> post-smoothing.
>
> Voxels of 1mm^3 ? Is you scanner for small animals ? Otherwise, for a
> first reconstruction, aim for voxels with their side equal to the crystal
> width.
>
> Finally, in a second step, you should consider including a normalization
> correction. There are already many discussions about that in the forum.
>
> For your first reconstructions with Gate simulated data, I would also
> recommend to include only true coincidences.
>
> Best
> Simon
>
>
> Le 12 avr. 2024 21:02, Dimitrios Thanasas <dthanas at uoa.gr> a écrit :
> Dear CASTOR-users,
>
> I simulate a simple PET ECAT system using GATE (V8.2 - I couldn't install
> CASTOR on version 9 or later).
> I simulated an orthogonal parallelepiped source (xlength:5 mm, ylength:1
> mm, zlength:10 mm).
> According to the guidelines, I have used castor-GATEMacToGeom to generate
> the geom file and castor-GATERootToCastor to create the Cdh and Cdf
> files.
> Then I ran the command
> castor-recon -df ecat_df.Cdh -opti MLEM -it 64:128 -dim 64.,64.,64. -vox
> 1.,1.,1. -dout myimages1
> after the termination of the reconstruction I get the attached image.
>
> I see something like circles in transverse1 plane instead of a orthogonal
> plane as I am expected. In coronal1 and sagital1 planes look like a slice
> to appear as I am expected but I think the images have few points. My
> GATE
> output root file includes 862000 coincidences.
> Is there something wrong with the reconstruction?
>
> Any help will be appreciated,
> Best regards,
> Dimitris
>
>