Question about Data Converter-Siemens Biograph mCT

General Discussions
1

Hi All,

I am trying to use the tool for the conversion of Siemens Biography mCT PET data to the CASToR data file format. In this tool, there are four mandatory options (output name, normalization sinogram, prompts sinogram and random sinogram). However, after using e7 toolbox, I do not have a random sinogram. How to solve this problem?

By the way, anyone using CASToR is from Australia or China?

Best wishes!
Yu

2

Hi Yu !

Is the original sinogram a prompt one, or a prompt-minus-delays ?
To know that, you may look into your sinogram to see if there are negative values or not.

Best
Simon

3

Dear Yu,

The random sinogram is mandatory to convert data into Castor format. When using e7tools, you can have access to the smoothed random sinogram (like the scatter sinogram…). It is named something like smoothed_rand_xx.s. This is directly the estimated random sinogram (with low variance) as computed by the approach developed by V. Panin.

Best,

Thomas

Nuclear Medicine

University hospital

Nantes, France

4

Hi Carlier,

Thanks for your reply. I checked my all files after using JSRecon from e7 toolbox. Did not find a file like smoothed_rand_xx.s. Would you mind sending me more details of how to get a smoothed random sinogram?

Best wishes!
Yu

Carlier Thomas <thomas.carlier@chu-nantes.fr> 于2019年4月10日周三 下午5:51写道：

5

Hi,

It is relatively straightforward. You should activate the flag -d when using e7_recon and of course specify the output directory.

Best

Thomas

6

Dear Simon,

I try to do random correction by delay window. After converting listmode to sinogram, I found there are negative valures (coincidence_sinogram - random_sinogram). I just set all negative value to 0 and found the noise from random does not reduce in sinogram which I add into one slice. Is it right to set 0?
Thanks for any reply!

Best,
Hannah

7

Hi Yu,

It’s a pity that I only saw this post now. I want to ask if you are still using CASToR. I am a user from China. Is there any place in China where I can discuss image reconstruction?

Best,
Hannah

8

Hi Hannah,

Random correction is usually done by modelling the addition of smoothed delays in the reconstruction rather by subtraction. This means that you need do have a smooth estimate of the randoms which is usually computed from delays and singles rates. See many publications from Vladimir Panin from Siemens.

In your case, if you do subtraction, because of inherent noise in the data, you end with negative values as you said. If you use the MLEM or OSEM algorithm, they cannot dela with negative values and what you do by truncating to 0 is mandatory. Obviously, it results into a positive bias that is inevitable.

I did not understand what you mean by “the noise from random does not reduce in sinogram which I add into one slice”.

Hope this helps

Simon

9

Dear Simon,

Thanks for your reply!

image
image527×912 128 KB

random coincidence in sinogram as shown in the second image(adding all slices), after subtraction, it looks empty in sinogram as shown in the third image. So I thought maybe coincidence in delay window has same position in sinogram but different in slices.

I try to search publications from Vladimir Panin about random correction, only found patents without any figures. It confuses me. I would like to find more details to figure out how to use delay window.

Thanks again!

Best,
Hannah

11

Dear Hannah,

How did you show the sinogram in this post? Did you show .cdf file? I want to read .cdf file and show the sinogram.

I would be grateful If you can help me.