pietz / brats-segmentation

Tumor Segmentation of the BRATS2015 dataset

BRATS2015 Tumor Segmentation (Work in Progress)

3D image segmentation using a U-Net-like CNN on the BRATS2015 dataset.

How To

1. Download the BRATS2015 dataset
2. Extract the archive in the directory of the Jupyter Notebook
3. Run the notebook

brats-segmentation's Issues

mask processing

Hello,

I do not understand why use these codes below for OT mask images, could you help me to illustrate it, please?

img[img == 4] = 1
img[img != 1] = 0

Why use img=4 not other numbers? is it related to T1c? I cannot find too much details about the dataset online, do you have some information about the dataset more detail?

Thanks!

About Weights

Hey @pietz how did you take the weights? When I try to run this without your weights the accuracy is pretty low. Did you change your weights by any means? Thanks

About f1_score

can you please show me how to run the "f1_score"?I try "f1_score(y[:50],pred)",but failed.I’m confused . thank you !

I need to see images in normal MRI image

Hi, when i run your code i'm get output images as you shown in sample. but i need those images in normal MRI Image view. Can you please tell me how can I make get images as i uploaded below?