CarinaNet: Automatic Detection of Carina and ETT in Chest X-rays

CarinaNet is a deep learning model for automatic detection of carina and ETT (endotracheal tube) in chest X-rays using a RetinaNet-based architecture. This implementation is based on the original CarinaNet model from USM-CHU-FGuyon/CarinaNet, adapted for use in the FactCheXcker project.

Model Description

Architecture: RetinaNet with ResNet backbone
Input Size: 640x640 pixels
Output: Coordinates for carina and ETT locations
Framework: PyTorch

Usage

Install the package:

pip install factchexcker-carinanet

Use the model:

import carinanet

# Simple prediction
result = carinanet.predict_carina_ett("chest_xray.jpg")
print(f"Carina: {result['carina']}")
print(f"ETT: {result['ett']}")

# Using the model class
model = carinanet.CarinaNetModel()
result = model.predict("chest_xray.jpg")

Model Performance

The model has been trained on curated medical imaging datasets and achieves high accuracy on clinical validation sets.

Citation

If you use CarinaNet in your research, please cite:

@article{guyon2023image,
  title={Image augmentation and automated measurement of endotracheal-tube-to-carina distance on chest radiographs in intensive care unit using a deep learning model with external validation},
  author={Guyon, Florian and others},
  journal={European Journal of Radiology},
  year={2023},
  volume={158},
  pages={110639},
  publisher={Elsevier},
  doi={10.1016/j.ejrad.2022.110639},
  url={https://pubmed.ncbi.nlm.nih.gov/36698191/}
}

@inproceedings{heiman2025factchexcker,
  title={FactCheXcker: Mitigating Measurement Hallucinations in Chest X-ray Report Generation Models},
  author={Heiman, Alice and Zhang, Xiaoman and Chen, Emma and Kim, Sung Eun and Rajpurkar, Pranav},
  booktitle={Proceedings of the Computer Vision and Pattern Recognition Conference},
  pages={30787--30796},
  year={2025}
}

License

This model is released under the Apache 2.0 License.

Disclaimer

This model is intended for research and educational purposes. It should not be used as the sole basis for clinical decisions.