Using ShiftNet

[taustudent]

Hello and thanks for your work! In the paper you mentioned the you used ShiftNer for the second stage, do you have the implementetion of this? Where it’s taking place, inside the PromptMR training loop or it’s separated training that starts only on the final results of PromptMR? Thanks

[hellopipu]

Hello and thanks for your work! In the paper you mentioned the you used ShiftNer for the second stage, do you have the implementetion of this? Where it’s taking place, inside the PromptMR training loop or it’s separated training that starts only on the final results of PromptMR? Thanks

It's separate training. We opted not to release that part of the code because the performance boost to the PromptMR results from separated training was minimal, and the process was time-consuming.

[taustudent]

Thanks for your answer. I want to reproduce this flow as final project in my Medical Image Processing course at university, so just to be sure, after I finish to train PromptMR and saved all final reconstructed images, I need to start new training of ShiftNet with on the train data of PromptMR?

[hellopipu]

Thanks for your answer. I want to reproduce this flow as final project in my Medical Image Processing course at university, so just to be sure, after I finish to train PromptMR and saved all final reconstructed images, I need to start new training of ShiftNet with on the train data of PromptMR?

yes!

[taustudent]

Thanks! And you train ShiftNet on the image space or k-space?

[hellopipu]

Thanks! And you train ShiftNet on the image space or k-space?

Image space. We saved the reconstructed images of the training set generated by PromptMR and then used them as input for ShiftNet during the second stage of training.