Skip to content

Latest commit

 

History

History
124 lines (99 loc) · 5.14 KB

BRAX.md

File metadata and controls

124 lines (99 loc) · 5.14 KB

BRAX Brazilian Chest X-ray

Dataset Information

The Brazilian Annotated X-ray (BRAX) dataset is an automatically annotated dataset consisting of chest X-ray studies from 24,959 patients in a large general hospital in Brazil. The dataset contains a total of 40,967 images.

14 Labels

Labels were automatically extracted from medical reports in Portuguese using NLP tools:

  1. No Finding: Value is 1 if no other label is present, except for support devices.
  2. Enlarged Cardiomediastinum
  3. Cardiomegaly
  4. Lung Lesion
  5. Lung Opacity
  6. Edema
  7. Consolidation
  8. Pneumonia
  9. Atelectasis
  10. Pneumothorax
  11. Pleural Effusion
  12. Pleural Other
  13. Fracture
  14. Support Devices

"master_spreadsheet.csv"

Each row in the dataset includes the following information:

  • PatientID: A unique identifier for the patient. As part of the de-identification procedure, patient IDs were randomly generated.
  • PatientSex: The sex of the patient. Enumerated values:
    • "M" for male
    • "F" for female
    • "O" for other
  • PatientAge: Patient's age is provided in 5-year age groups. Patients aged 85 or older are classified as "85 or more."
  • AccessionNumber: A DICOM identifier for the study. This was randomly generated as part of the de-identification process.
  • StudyDate: A fictitious date for the study.
  • Labels: Columns indicating the presence of the 14 labels. The code "1" represents positive, "0" represents negation, and "-1" indicates uncertainty.
  • ViewPosition: The radiographic view associated with the patient's position. Defined terms:
    • AP - Anterior/Posterior
    • PA - Posterior/Anterior
    • LL - Left Lateral
    • RL - Right Lateral
    • RLD - Right Lateral Decubitus
    • LLD - Left Lateral Decubitus
    • RLO - Right Lateral Oblique
    • LLO - Left Lateral Oblique
  • Rows: The size (number of pixels) along the vertical axis of the image matrix.
  • Columns: The size (number of pixels) along the horizontal axis of the image matrix.
  • Manufacturer: An index representing the CT scanner's manufacturer. Manufacturer names are coded as integers to conceal their identities while still allowing future research on possible biases related to vendor/machine settings.

Dataset Meta Information

Dimensions Modality Task Type Anatomical Structures Number of Categories Data Volume File Format
2D X-Ray Unsupervised Representation Learning, Classification Heart, chest and lungs 14 24,959 cases, 19,351 patients, 40,967 images PNG, DICOM

Resolution Details

Dataset Statistics size
min 1024x1082
median 1024x1082
max 1024x1082

Visualization

File Structure

Anonymized_DICOMs
├── id_00082e3a-ec11c281-24a79518-35d3cc78-22432fb1
│   ├── Study_09342613.22970294.40563343.35634289.53163857
│   │   ├──Series_34523850.21768222.07508551.49190893.14603932
│   │   │   ├── image-48219538-15808688-10728535-52591088-74513595.dcm
│   │   ├──Series_46177599.95157937.50203011.63555832.78161828
│   │   │   ├── image-48219538-15808688-10728535-52591088-74513595.dcm
│   ├── Study_51027964.83117427.20948980.39828954.71003607
│   │   ├──Series_57104384.74837822.26263330.97688944.88328246
│   │   │   ├── image-48651870-23127024-63651831-17193122-94277772.dcm
│   │   ├──Series_72993604.79060724.14705971.37953714.05369399
│   │   │   ├── image-08788867-77959894-95405066-47915205-10581326.dcm
master_spreadsheet.csv

Authors and Institutions

Eduardo Pontes Reis (Hospital Israelita Albert Einstein)

Source Information

Official Website: https://physionet.org/content/brax/1.1.0/

Download Link: https://physionet.org/content/brax/1.1.0/

Article Address: https://physionet.org/content/brax/1.1.0/

Publication Date: 2022-06

Citation

@article{reis2022brax,
  title={BRAX, Brazilian labeled chest x-ray dataset},
  author={Reis, Eduardo P and De Paiva, Joselisa PQ and Da Silva, Maria CB and Ribeiro, Guilherme AS and Paiva, Victor F and Bulgarelli, Lucas and Lee, Henrique MH and Santos, Paulo V and Brito, Vanessa M and Amaral, Lucas TW and others},
  journal={Scientific Data},
  volume={9},
  number={1},
  pages={487},
  year={2022},
  publisher={Nature Publishing Group UK London}
}

Original introduction article is here.