This README file records the operation of data preprocessing, 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 |