Home

STIR (Software for Tomographic Imaging Reconstruction) is Open Source software for use in tomographic imaging. Its aim is to provide a Platform-independent Object-Oriented framework for all data manipulations in tomographic imaging.

This page is a place-holder. The actual wiki is at

https://github.com/UCL/STIR/wiki/

Discussions entered here will be ignored. Please use the mailing lists and/or Discussions on GitHub.

[Amir]

Hello Kris,

I have a few questions related to PET image reconstruction using OSMAPOSL in STIR 2.4. I am reconstructing a Phillips PET image that is pre-corrected for random coincidences at the acquisition unit and would like to correct for the effects of normalization, attenuation, and scatter using STIR. I’ve tested the following 3 approaches to reconstruction, as mentioned in a previous post, in order to obtain the best quality image:

1. Pre-correction of input sinogram for the effects of normalization, scatter (using Simulated Single Scatter), and attenuation followed by reconstruction with a post filter.

2. Attenuation correction of input sinogram by feeding attenuation correction factors (ACF) into the Bin Normalization procedure and computing the additive sinogram according to the following formula: A = r + s/(NACF) from this thread, where:
   • A = additive sinogram
   • r = random coincidences
   • s = scatter computed according to Single Scatter Simulation
   • N = normalization sinogram
   • ACF = attenuation correction factors
   Post-filtering was applied here for best results.

3. Attenuation correction of input sinogram by feeding ACF into the Bin Normalization procedure and computing the additive sinogram according to the following formula: A = (rN + s)ACF – from STIR manual – where the variables are as above. Post-filtering with 23 subsets produced a blank image, while inter-filtering produced decent results but with correlated noise.

In approaches 2 and 3, correcting for normalization by either multiplying the input sinogram by N or inputting N into Bin Normalization produced a blank image. This is restricting me from correcting for normalization in methods 2 and 3.

I have the following questions:

1.  Which is theoretically the best approach?
2.  Which is the correct formula for computing an additive sinogram? 
3.  Why does a choice of inter vs. post filtering affect whether or not an image is produced? 
4.  Why is method 1 producing the best image when according to the STIR manual and FAQ posts approach 3 should produce the best image?
5.  Why do I get a blank image when I try to correct for norm? If this is due to non-convergence, how may I overcome this?

I’ve attached my par files and corresponding reconstructed images (jpeg) of the best results from each approach.

Thank you in advance for your insight.

Regards,

Amirreza Abbasnejad
NSHealth | Diagnostic Imaging Research

[Shahzad Ali]

Dear Amir Can you send me the macros for the reconstructions. I am new and do not understand the procedure properly.
I will be thankful to you alot. My email ID is "s_alikhan92@yahho.com".

[Itzel Gracia]

Hi!
I'm having problems installing the library, it doesn't send out any error messages but I don't seem to be able to compile or even run a single command on terminal.

Am i doing something wrong? Do you have any advice or suggestions?

Thanks in advance.