10-fold cross-validation on the training set.

[wenhankang]

Excuse me,'ten models were trained via 10-fold cross-validation on the training set' In the paper.But I only see the last ten models saved and a baseline, not the ten cross-validation runs in the code.Could you please help me explain how to carry out ten times of cross verification?

[wenhankang]

@ellisdg

[ellisdg]

To do 10 fold cross validation, you need to set the training and test subjects in a separate config file for each fold. You can see the cross validation folder from the autoimplant project for an example of 4 fold cross validation. The trianing can then be run on each config file separately.

[wenhankang]

要进行 10 折交叉验证，您需要在单独的配置文件中为每个折设置训练和测试主题。您可以从 autoimplant 项目中查看交叉验证文件夹，以获取 4 折交叉验证的示例。然后可以分别在每个配置文件上运行 trianing。

'm glad to receive your reply. Could you help me solve how to use the ten model U-Net ensemble on the BraTS 2020 test set.

[ellisdg]

Here are the configuration files I used to train the models:

1. https://github.com/ellisdg/fCNN/blob/master/data/brats_config_fold0.json
2. https://github.com/ellisdg/fCNN/blob/master/data/brats_config_fold1.json
3. https://github.com/ellisdg/fCNN/blob/master/data/brats_config_fold2.json
4. https://github.com/ellisdg/fCNN/blob/master/data/brats_config_fold3.json
5. https://github.com/ellisdg/fCNN/blob/master/data/brats_config_fold4.json
6. https://github.com/ellisdg/fCNN/blob/master/data/brats_config_fold5.json
7. https://github.com/ellisdg/fCNN/blob/master/data/brats_config_fold6.json
8. https://github.com/ellisdg/fCNN/blob/master/data/brats_config_fold7.json
9. https://github.com/ellisdg/fCNN/blob/master/data/brats_config_fold8.json
10. https://github.com/ellisdg/fCNN/blob/master/data/brats_config_fold9.json

Train a model using each configuration file, this gives you 10 models. If you want to do the cross-validation, predict the "validation" subjects for each model according to its configuration file. If you want to predict the test dataset, use each model to predict each of the test subjects and average the result.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. If you are still wanting followup to this issue, please ping the thread by leaving a comment. You may also contact david.ellis@unmc.edu with questions.