search

xiawj-hub / CTLIB

A lib of CT projector and back-projector based on PyTorch
MIT License
39 stars 6 forks source link

options problem - dataset_AAPM #15

Open shuyifan1024 opened 4 months ago

shuyifan1024 commented 4 months ago

Hello,

I meet a trouble in options parameter. This is dataset description: https://qtim-challenges.southcentralus.cloudapp.azure.com/competitions/1/ This is reconstruction function code: https://github.com/xcist/example/blob/main/AAPM_datachallenge/AAPMRecon.py This code contains some key parameters of this dataset.

options = np.array([1000, 900, 512, 512, 0.0078125, 0.6/400, 0, 2*np.pi/1000, 5.5, 4, 0, 1])

I used this set of parameters into CTLIB fbp. But the reconstruction came out wrong. Could you help me check this options? untitled

Thank you so much. best

xiawj-hub commented 4 months ago

According to the reconstruction, there is probably a data flipping problem, originating from different coordinating system of different scanners. You can try to flip the projection data, I don't know which dimension you need to flip, maybe one of them or both. If there is still slight artifacts after flipping, it may be from the detector offset, try to set offset parameters.

xiawj-hub commented 4 months ago

Flipping, not rotating, probably the detector dimension, you can try first.

From: 小李努力再努力 @.> Sent: Thursday, April 18, 2024 9:56 PM To: xiawj-hub/CTLIB @.> Cc: xiawj-hub @.>; Comment @.> Subject: Re: [xiawj-hub/CTLIB] options problem - dataset_AAPM (Issue #15)

According to the reconstruction, there is probably a data flipping problem, originating from different coordinating system of different scanners. You can try to flip the projection data, I don't know which dimension you need to flip, maybe one of them or both. If there is still slight artifacts after flipping, it may be from the detector offset, try to set offset parameters.

rotate direction？ Can it be solved by just set dAng to negative while keep Ang0 to 0？

— Reply to this email directly, view it on GitHub https://github.com/xiawj-hub/CTLIB/issues/15#issuecomment-2065610119 , or unsubscribe https://github.com/notifications/unsubscribe-auth/AJRVWQMKBQNFIL4RZEMVUYLY6B2STAVCNFSM6AAAAABGOJAE2KVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDANRVGYYTAMJRHE . You are receiving this because you commented. https://github.com/notifications/beacon/AJRVWQNXNN3AFHFTF6DK3Z3Y6B2STA5CNFSM6AAAAABGOJAE2KWGG33NNVSW45C7OR4XAZNMJFZXG5LFINXW23LFNZ2KUY3PNVWWK3TUL5UWJTT3D22YO.gif Message ID: @. @.> >

shuyifan1024 commented 4 months ago

According to the reconstruction, there is probably a data flipping problem, originating from different coordinating system of different scanners. You can try to flip the projection data, I don't know which dimension you need to flip, maybe one of them or both. If there is still slight artifacts after flipping, it may be from the detector offset, try to set offset parameters.

Thank you so much for your help. I adjust the flipping. It works.