---
license: gpl-2.0
task_categories:
- image-segmentation
tags:
- medical-imaging
- ct
- kidney
- segmentation
---

# CT2USforKidneySeg

CT (source-domain) slices with kidney segmentation masks used in
**Song et al., "CT2US: Cross-modal transfer learning for kidney
segmentation in ultrasound images with synthesized data."**
*Ultrasonics* 122 (2022) 106706. DOI:
[10.1016/j.ultras.2022.106706](https://doi.org/10.1016/j.ultras.2022.106706).

Mirror of the public Kaggle release
[siatsyx/ct2usforkidneyseg](https://www.kaggle.com/datasets/siatsyx/ct2usforkidneyseg).

## Contents
- **4586** paired samples at 256x256, single split `train`.
- `image`: grayscale CT slice (PNG, 8-bit).
- `mask`: kidney segmentation mask (PNG, 8-bit; mostly binary with minor
  anti-aliasing artifacts; threshold at 127 to recover the binary label).
- `id`: original Kaggle filename stem (non-contiguous, drawn from multiple CT volumes).

## License
GPL-2.0 (per the original Kaggle dataset).

## Citation
```
@article{song2022ct2us,
  title   = {CT2US: Cross-modal transfer learning for kidney segmentation in ultrasound images with synthesized data},
  author  = {Song, Yuxin and Zheng, Jing and Lei, Long and Ni, Zihan and Zhao, Baoliang and Hu, Ying},
  journal = {Ultrasonics},
  volume  = {122},
  pages   = {106706},
  year    = {2022},
  doi     = {10.1016/j.ultras.2022.106706}
}
```