The rationality of automatic configuration strategies for nnWNet

[regedit818]

Dear authors, hello. Thank you for publishing an excellent paper and for also open-sourcing your code. The number of z-axis values in our dataset is far less than that in the dataset you used, which also led to the original configuration being unusable and resulted in a poor outcome.

I modified both the upsampling and downsampling strategies to automatically use the strides set by nnU-Net. Is this reasonable? In addition, what if the number of layers and channel is also taken from nnU-Net’s configuration?

[Yanfeng-Zhou]

Thank you for your interest in our work!
Any configuration of the model architecture within the nnUNet framework will not affect the nnWNet architecture.
Only configurations unrelated to the model architecture, such as dataset preprocessing, data augmentation, and training hyperparameters, will be applied to nnWNet.

[regedit818]

Thanks for your response.