akhanss / brats-2020

A complete pipeline for BraTS 2020

Hi there 👋

• 🔭 I’m currently working on deep learning-based medical image analysis tasks.

brats-2020's Issues

Issue in metrics.py while running on Colab

Hi!
I am running this project on Google Colab using GPU. Everything works perfectly until it reaches the metrics.py file. Here it gives the following error:

I assume this is due to an upgrade in the Tensorflow Library. Any suggestions on how to fix it?

Saved model share request

Thank u for uploading this code but can you please share save model and weights

ImportError: cannot import name 'Deconvolution3D'

The error in data.py

Thank you for sharing! But I find that always has ERROR like
File "/home/server/zMap/BraTS-2020-master-1/src/unet3d/data.py", line 53, in write_data_to_file
n_channels = len(training_data_files[0]) - 1
IndexError: list index out of range

ET, WT, TC and NCR/NET, ED, ET relations

I am a little confused about the predicted labels.

As described in the data description, the segmentations label annotations are the GD-enhancing tumor (ET — label 4), the peritumoral edema (ED — label 2), and the necrotic and non-enhancing tumor core (NCR/NET — label 1). However, the results you showed are in ET, WT, TC.

Using the training data, we can only infer the label annotations, which are NCR/NET, ED, ET. There should be some conversion between these annotations and ET, WT, TC, but I didn't see any implementation on this.

Could you please explain this confusion?

from where i can get validation data for BraTS 2020. Job request is only giving access to training data

memory dump while starting training

hello
I was trying to run your code (training) - thanks for sharing-, exactly as described and when train_model starts I get a memory dump. I played with different image sizes and with patches, no change. I am running on ubuntu 1.8, tf 1.15 with a RTX2080ti. This setting works with other nets. Any help appreciated

Issue in metrics.py function weighted_dice_coefficient

Thanks for the sharing. I am facing issue in function "weighted_dice_coefficient" of metrics.py.
TypeError: Input 'y' of 'Mul' Op has type float32 that does not match type int8 of argument 'x'.

I printed the y_true and y_pred and found that
printing y_true in console = Tensor("IteratorGetNext:1", shape=(None, None, None, None, None), dtype=int8)
printing y_pred in console = Tensor("model/activation/Sigmoid:0", shape=(None, 3, 128, 128, 128), dtype=float32)
printing y_true.shape in console = (None, None, None, None, None)
printing y_pred.shape in console = (None, 3, 128, 128, 128)
Please see the error screenshot

Can you please help in this regard