About BraTS sample result

[springbreeze7111]

Thanks to this effort, I tried to reproduce the BRATS2021 dataset, and I modified the saved file to NII format. I wondered why the final sample result was an MRI image without a mask, how do I get the result of the generated prediction mask to compute the metric？

[FelixzeroSun]

Can I ask u how do you solve this problem

[springbreeze7111]

Can I ask u how do you solve this problem

I modify the return value saved by the sample script, specifically in line 136 of segmentation_sample. I change the sample to cal, which is what looks like the predicted mask, and then modify the code behind it to save it in .nii format, and then go back and calculate the metrics. This is my current idea and I'm trying it all over again.

[FelixzeroSun]

Can I ask u how do you solve this problem

I modify the return value saved by the sample script, specifically in line 136 of segmentation_sample. I change the sample to cal, which is what looks like the predicted mask, and then modify the code behind it to save it in .nii format, and then go back and calculate the metrics. This is my current idea and I'm trying it all over again.

Thanks for your reply!
Do you mean this line from the first screenshot to the second one?
Did you do any other modifications?
Thanks in advance!

[springbreeze7111]

Can I ask u how do you solve this problem

I modify the return value saved by the sample script, specifically in line 136 of segmentation_sample. I change the sample to cal, which is what looks like the predicted mask, and then modify the code behind it to save it in .nii format, and then go back and calculate the metrics. This is my current idea and I'm trying it all over again.

Thanks for your reply! Do you mean this line from the first screenshot to the second one? Did you do any other modifications? Thanks in advance!

Right, as modified in your second picture. My other modifications are mainly for the file saving format changes in the source code. For example, at the end of the sample script you can see that the author saved the output as a jpg, which I changed to save it in the same .nii format as the original dataset. For the previous modification of the return you can see #106 (comment) . This discussion has shown what all the return values represent.

[FelixzeroSun]

Thanks for your help, and I finally figure it out. You made my day bro!