search

Kaiseem / DAR-UNet

[JBHI2022] A novel 3D unsupervised domain adaptation framework for cross-modality medical image segmentation
Apache License 2.0
41 stars 5 forks source link

Data preprocess #1

Closed det-tu closed 2 years ago

det-tu commented 2 years ago

Hi, i wonder in step1 of data processing, how to convert multiple 3d nii files to npy? Is there any requirements? Or can you please release the code of data processing?

Thanks!

Kaiseem commented 2 years ago

Hi, for your question, you can load the array through simpleITK library from 3D nii files and save them as npy, e.g., np.save.

Meanwhile, I have detailed all the necessary and general steps to prepare the datasets, similar with TransUNet, you can also refer to the paper for more detials, e.g., specific spacing for each datasets. And there is no special requirement here to preprocess them.

det-tu commented 2 years ago

Thanks for your answer! For target image in CrossModa, the labels are lacked so can't generate label.npy, how to complete the steps in data preprocessing?

Kaiseem commented 2 years ago

You are right about it, for the target domain, the annotation is not accessible. So you have to generate the prediction on the target domain with the trained segmentor, and then evaluate your method by submitting the prediction to the official website of CrossMoDa. I don't have the annotation too, so I got the results by submitting the predictions.