JustlfC03 / TC-DiffRecon

[ISBI 2024] TC-DiffRecon: Texture coordination MRI reconstruction method based on diffusion model and modified MF-UNet method
MIT License
16 stars 1 forks source link

How do I get the reconstructed MRI images? #2

Open lzl2040 opened 3 months ago

lzl2040 commented 3 months ago

In image_sample_complex_duo.py, the amplitude of the frequency is saved as a picture. Could you tell me how to get the reconstructed MRI image? Is it the inverse Fourier transform of sample2? Thank you!

JustlfC03 commented 3 months ago

I don't quite understand what you mean, what you get after model sampling image_sample_complex_duo.py is the reconstructed MRI image. Are you talking about k-space images? k-space image is obtained by MRI after inverse Fourier transform.

lzl2040 commented 3 months ago

I don't quite understand what you mean, what you get after model sampling image_sample_complex_duo.py is the reconstructed MRI image. Are you talking about k-space images? k-space image is obtained by MRI after inverse Fourier transform.

Thank you for your reply. I have run 20,000 steps with batch size 2 and run image_sample_complex_duo.py, but the reconstructed MRI image is as follows: file1000033_16 (1)

Is this because the training steps are not enough? I find that the loss is already very small.

JustlfC03 commented 3 months ago

I don't think so. I think you've entered the data in the wrong format.

lzl2040 commented 3 months ago

When I downloaded the fastMRI dataset from the OneDrive, I followed readme.md to first run data_process.py, then run image_train.py. 2

JustlfC03 commented 3 months ago

I know where you are wrong, you should have misunderstood my data_process.py code, the comment code misled you. I now re-update the code, you can re-download this code to try again.

lzl2040 commented 3 months ago

Thank you so much!