search

ivy9092111111 / TT-U-Net

TT U-Net: Temporal Transformer U-Net for motion artifact reduction in dynamic cardiac CT
19 stars 2 forks source link

can you provide more information on the PAD operation? #5

Open medicatedfizz opened 3 months ago

medicatedfizz commented 3 months ago

Hello. I have sent you an email regarding the identity of the files in the PAD operation. I had not received any answer, so I ask the questions in your github as well.

Here are my questions:

  1. I'm confused about what the ‘utils/rpc.mat’ file corresponds to in accordance with the notation in the paper. My first guess is that it corresponds to the s_bar value (mean of intra + inter in XCAT). Is this correct?

  2. I'm confused about the ‘utils/volportion’.mat file as well. My guess is that it is the label2/(label1+label2) value for the 20 phases of the 3D intra mean of XCAT data, is this correct?

  3. Lastly, this is the most important: I cannot figure out meaning of files under the ‘shape/‘ directory. I'm guessing they are obtained from XCAT data, but I'm not sure how to acquire those values from XCAT data. Could you explain what each values correspond to using the notations in the paper, and how to obtain them from the XCAT simulation? Or can you upload some more details about the PAD pipeline in the TT U-Net github repository?

ivy9092111111 commented 3 months ago

Thank you for your attention!

Sorry for the delay. I just find that your e-mail was misidentified as a spam. That was not a smart system.

  1. The rpc is my abbreviation for reference point cloud. All the point cloud model should be align with it and then to determine the pointwise correspondence.

  2. Yes, that's correct.

  3. Sorry I can not provide too many details about the way we establish the 4D SSM. It's not about the technique, but we do not have the access to share either the XCAT phantoms or the real 4D cardiac CT images. As you may have noticed, we have withdrew the google drive link recently (the PAD_pipeline_github.zip file you have downloaded). The 4D SSM is a well developed technique, and we recommend the following reference: C. Hoogendoorn, “A high-resolution atlas and statistical model of the human heart from multislice CT”. M. Unberath, "Open-source 4D statistical shape model of the heart for X-ray projection imaging". M. Wilms, "A 4D Statistical Shape Model for Automated Segmentation of Lungs with Large Tumors".

I hope you find them informative and helpful!

medicatedfizz commented 2 months ago

Thank you for the detailed reply. Your answer was very helpful, especially for the third question. I will look into the sources you have provided.

But I am still a bit confused about what the rpc.mat file is. I understand it is the reference cardiac point cloud, and is used for applying linear registration to the MMWHS data.

However, I am confused about where the rpc.mat file is derived from.

  1. Is it derived from the XCAT data? If so, is it the mean shape of all people and phases(inter + intra mean)?

  2. Can it be used for every data in the MMWHS dataset, or is it specifically designed for the img3.mat file only?

ivy9092111111 commented 2 months ago
  1. Yes, it is derived from the XCAT data. I guess I just randomly selected one as the reference shape. Maybe it is better to use the mean shape.
  2. Actually it can be used for any cardiac CT image, not only for the MMWHS dataset.
ivy9092111111 commented 2 months ago
  1. Yes, it is derived from the XCAT data. I guess I just randomly selected one as the reference shape. Maybe it is better to use the mean shape.
  2. Actually it can be used for any cardiac CT image, not only for the MMWHS dataset.
medicatedfizz commented 2 months ago

Thank you for answering all my questions. All my questions are resolved for now, but I might leave more questions in the future.