ykumards/open-i

Chest X-rays (Indiana University)

Copy of the kaggle dataset: https://www.kaggle.com/datasets/raddar/chest-xrays-indiana-university created by raddar

---

Open access chest X-ray collection from Indiana University

Original source: https://openi.nlm.nih.gov/

Original images were downloaded in raw DICOM standard. Each image was converted to png using some post-processing:

top/bottom 0.5% DICOM pixel values were clipped (to eliminate very dark or very bright pixel outliers)
DICOM pixel values scaled linearly to fit into 0-255 range
resized to 2048 on shorter side (to fit in Kaggle dataset limits)

Metadata downloaded using available API (https://openi.nlm.nih.gov/services#searchAPIUsingGET)

Each image classified manually into frontal and lateral chest X-ray categories.

License: Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0)

---

Usage

The lateral and frontal images of each uid are grouped together in an example. The images are stored as bytes, and can be loaded to PIL Image using the following method

def load_image_from_byte_array(byte_array):
    return Image.open(io.BytesIO(byte_array))

Cite

Please site the original source of the dataset.

@article{demner2016preparing,
  title={Preparing a collection of radiology examinations for distribution and retrieval},
  author={Demner-Fushman, Dina and Kohli, Marc D and Rosenman, Marc B and Shooshan, Sonya E and Rodriguez, Laritza and Antani, Sameer and Thoma, George R and McDonald, Clement J},
  journal={Journal of the American Medical Informatics Association},
  volume={23},
  number={2},
  pages={304--310},
  year={2016},
  publisher={Oxford University Press}
}