search

med-air / DLTTA

[IEEE TMI'22] DLTTA: Dynamic Learning Rate for Test-time Adaptation on Cross-domain Medical Images
16 stars 2 forks source link

About prostate segmentation. #2

Closed Asagami-Fujino closed 2 years ago

Asagami-Fujino commented 2 years ago

Thanks for your work. I'm trying to reproduce results on Prostate datasets. But I really don't understand the sentence 'Patches with size of 80x80x80 were cropped as the network inputs.' in the paper. The shape of the MRI volumes is aound [20, 384, 384], which is greatly different from the network input. I'm wondering how to crop the MRI volums into 80x80x80? Looking forward to your reply!

HongzhengYang commented 2 years ago

Thanks for your interest. For the prostate dataset, the resolution (in/through plane) is quite different among the six sites. So we conducted re-spacing before cropping the patch. After re-spacing, the shape of the MRI volumes is around [256, 256, 128].

Asagami-Fujino commented 2 years ago

Thanks @HongzhengYang, the problem is solved, thanks for your prompt reply :-)

indranarendra commented 1 year ago

So we conducted re-spacing before cropping the patch. After re-spacing, the shape of the MRI volumes is around [256, 256, 128].

Can you please explain how this re-spacing is done? I'm also trying to reproduce these results. So, it will be helpful for me if you share the code for this.

HongzhengYang commented 1 year ago

Hi, thanks for your interest. I have uploaded the re-spacing code. (https://drive.google.com/drive/folders/14Joa9CstJYkyjNLZwU6bmy4DDgb3XgzE?usp=sharing)