clinical-super-mri.

MRI super-resolution (work in progress)

Table of content

1. Installation
2. Dataset
3. Train model
4. Monitoring and Visualization
5. Prediction

1. Installation

Create new conda environment supermri
```
conda create -n supermri python=3.8
```
Activate supermri environment
```
conda activate supermri
```
Run installation script
```
sh setup.sh
```
Restart conda environment after installation.

2. Dataset

See data/RAEDME.md for the structure of the dataset and how to use the pre-processing script.

3. Train model

train.py is the main file to train different models and datasets, it also supports tensorboard monitoring, checkpoint saving and mixed precision training. The following is an example on how to train a UNet on MDS.
Activate supermri conda environment
```
conda activate supermri
```

The following command train a UNet on MDS for 100 epochs, and store model checkpoint and summary to runs/001_unet_mds

python train.py --input_dir data/dataset/MDS --output_dir runs/001_unet_mds --model unet --num_filters 64 --patch_dim 32 --n_patches 1000 --merge_scan_type --batch_size 32 --epochs 100 --mixed_precision

To see all the input arguments, use the --help flag.
```
python train.py --help
```

4. Monitoring and Visualization

To monitor training performance and up-sampled results stored in --output_dir, we can use the following command
```
tensorboard --logdir runs/001_unet_mds
```
By default, TensorBoard starts a local server at port 6006, you can check the TensorBoard summary by visiting localhost:6006.

5. Prediction

After training the model with checkpoint saved at --output_dir, you can use predict.py to up-sample any scans in .mat format.

For instance, if scans needed to be up-sampled are stored in data/test_scans then we can use the following command

python predict.py --input_dir data/test_scans --output_dir runs/001_unet_mds --upsample_dir data/test_scans/upsample

The up-sampled scans would be stored under --upsample_dir with the same name as the original scan.
Note that scans have keys: {FLAIRarray, T1array, T2array} in the .mat file.

bryanlimy/sr-mri