PAXRay Dataset

PAXRay (Projected dataset for the segmentation of Anatomical structures in X-Ray data) is a synthetic dataset designed to facilitate research on anatomical segmentation in X-Ray imagery. The dataset contains projections of the RibFrac CT dataset onto a 2D plane to simulate realistic X-Ray data. It includes multi-label segmentation masks for fine-grained and hierarchical anatomical labeling.

For an extended version of this dataset, check out PAXRay++.

For models for anatomy segmentation of Chest X-Rays, check out CXAS.

---

Dataset Overview

• Images: 880 synthetic X-Ray-like images, generated from CT projections.
• Views: Both lateral and frontal views are included.
• Segmentation Labels:
  • Fine-grained: 92 individual anatomical labels.
  • Super-classes: Combined into 166 hierarchical labels.
• Purpose: Enabling segmentation of anatomical structures in X-Ray imagery for tasks such as:
  • Medical imaging analysis.
  • Transfer learning to real-world X-Ray segmentation.
  • Understanding anatomical structures for report generation.

---

Usage

Dataset Access

The dataset is not directly hosted here but is used with the provided tools and repositories for loading and applying models.

Loading the Dataset

For dataset loaders and utilities to work with PAXRay and other similar datasets, refer to the 2DAnatomyDatasets repository.

Model Application

For models trained and applied to this dataset, such as those for segmentation tasks, refer to the ChestXRayAnatomySegmentation repository.

---

Citation

If you use this dataset in your work, please cite:

@inproceedings{Seibold_2022_BMVC,
  author    = {Constantin Marc Seibold and Simon Reiß and M. Saquib Sarfraz and Matthias A. Fink and Victoria Mayer and Jan Sellner and Moon Sung Kim and Klaus H. Maier-Hein and Jens Kleesiek and Rainer Stiefelhagen},
  title     = {Detailed Annotations of Chest X-Rays via CT Projection for Report Understanding},
  booktitle = {33rd British Machine Vision Conference 2022, {BMVC} 2022, London, UK, November 21-24, 2022},
  publisher = {{BMVA} Press},
  year      = {2022},
  url       = {https://bmvc2022.mpi-inf.mpg.de/0058.pdf}
}

---

Acknowledgments

PAXRay leverages projections from the RibFrac CT dataset and benefits from extensive annotation efforts. This dataset has been designed to advance research in medical imaging, especially for tasks involving segmentation and understanding of chest X-Rays.

For further information or inquiries, please contact Constantin Marc Seibold.