Data preprocessing
June 26, 2024 ยท View on GitHub
Synapse dataset
The following desription provides the operation of data preprocessing for Synapse dataset, corresponding to the operations in preprocess_data.py
- Please download the Synapse dataset from the official Synapse website. Convert them to numpy format, clip the value range to [-125,275], normalize the 3D image to [0-1], extract the 2D slices from the 3D image for training and store the 3D images as h5 files for inference.
- According to the data description of the Synapse dataset, the map of the semantic labels between the original data and the processed data is shown below.
| Organ | Label of the original data | Label of the processed data |
|---|---|---|
| spleen | 1 | 1 |
| right kidney | 2 | 2 |
| left kidney | 3 | 3 |
| gallbladder | 4 | 4 |
| liver | 6 | 5 |
| stomach | 7 | 6 |
| aorta | 8 | 7 |
| pancreas | 11 | 8 |
LA dataset
The following desription provides the operation of data preprocessing for LA dataset:
- Please download the LA dataset and related list files from https://github.com/yulequan/UA-MT/tree/master/data.
- Run preprocess_LA.py.
PROMISE12 dataset
The following desription provides the operation of data preprocessing for PROMISE12 dataset:
- Please download the preprocessed PROMISE12 dataset from https://drive.google.com/file/d/1Xaj9MuYxaqzA28rq8E3VZe6A9rZI5SvU/view?usp=share_link.
- Run preprocess_PROMISE12.py.