Segmentation in Medical Image
This project aims at providing deep learning models in medical image segmentation.
Dependencies
- pytorch
- torchvision
- tensorboard_logger
- SimpleITK (optional)
Usage
Prepare the dataset
-
Train
python train.py
Customize
- Prepare your own dataset and write the
torch.nn.Dataset
for specific data in folder data - Design your network model in folder model
- Design the evaluation procedure in folder evaluate
- Write your own configuration file in folder config
- Modify the
dataset_name
to your dataset in train.py - Finally, run
python train.py
Experiment
-
ICH210
Dice Coeff (Std.) | IoU (Std.) | Sensitivity (Std.) | Specificity (Std.) | |
---|---|---|---|---|
U-Net | 82.69 (2.17) % | 70.55 (3.13) % | 75.99 (3.05) % | 90.76 (1.72) % |
-
BraTS19
License
Only academic study