Dr.Raum

Dr. Raum is a brain tumor segmentation project using the UNETR model. More details are on my blog.

How to Train

1. Install packages

$ pip install -r requirements.txt

2. Create kaggle token
   Download the kaggle token and put it in the Dr.Raum folder

{"username":"?????","key":"????????????????????????????"}

3. Download dataset

$ mkdir -p ~/.kaggle
$ cp kaggle.json ~/.kaggle/
$ chmod 600 ~/.kaggle/kaggle.json
$ kaggle datasets download awsaf49/brats20-dataset-training-validation
$ unzip -qq /content/brats20-dataset-training-validation.zip

4. Set config.py values and Train

$ python3 train.py

Dice Coefficient

# Train
Loss = 0.2376
Dice_Score = 0.7878

# Validation
Loss = 0.2280
Dice_Score = 0.7977

Reference

Web

1. Preprocess : https://www.kaggle.com/code/zeeshanlatif/brain-tumor-segmentation-using-u-net
2. F1 Score: https://velog.io/@jadon/F1-score%EB%9E%80
3. Dice Loss: https://attagungho.tistory.com/11#index
4. UNETR : https://kimbg.tistory.com/33

Paper

1. Transformer : https://arxiv.org/abs/1409.0473
2. 3D U-net : https://arxiv.org/abs/1606.06650
3. Vision Transformer : https://arxiv.org/abs/2010.11929
4. UNETR : https://arxiv.org/abs/2103.10504