Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

NAC PET to AC PET #1

Open
sarabanoo opened this issue Apr 27, 2023 · 3 comments
Open

NAC PET to AC PET #1

sarabanoo opened this issue Apr 27, 2023 · 3 comments

Comments

@sarabanoo
Copy link

Hi, can I use this codes ( cycle-GAN or U_net ) for another dataset. I want to convert NAC PET to AC PET ?

@RawthiL
Copy link
Owner

RawthiL commented May 1, 2023

I think that you can adapt the code to go directly from NAC-PET to AC-PET. There is some literature about it.
You can modify this code to achieve that.
Keep in mind that the result can contain hallucinated structures. This was one of the reasons that we had to avoid going directly from NAC-PET to AC-PET. We wanted mainly to estimate the attenuation image to then improve the effectiveness of simultaneous emission and attenuation reconstruction by providing a high quality prior.

P.S.: I have finished the thesis since my last update to this code, you might want to check it out here.

@sarabanoo
Copy link
Author

thank's RawthiL . Do you have an opinion about which neural network is better for going directly from NAC-PET to AC-PET?
The dataset that I have has scattered corrections applied to it and only the attenuation correction has not been applied to dataset.

@RawthiL
Copy link
Owner

RawthiL commented May 13, 2023

You will need to review current bibliography, but as of 2020 the best approaches were to use GAN (like this topologies)
You can start looking references to this work to see if there is any new one.
https://pubmed.ncbi.nlm.nih.gov/31869826/

If you are looking for released/public code for the task of translating NAC-PET to AC-PET , I'm not aware of any. The experience I had trying to reach other authors for code releases or testing their models on new data was never successful.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants