CAMUS - 2-D Echocardiographic Ultrasound Dataset

This dataset is a zea-format (HDF5) conversion of the CAMUS dataset for multi-structure segmentation in 2-D echocardiography.

Property	Value
Modality	2-D transthoracic echocardiography
Patients	500
Views	2-chamber (2CH) and 4-chamber (4CH) apical
Splits	train (1-400), val (401-450), test (451-500)

Conversion

This dataset was downloaded, converted to zea format, and uploaded using the zea data converter:

python -m zea.data.convert camus <src> <dst> --download

Dataset structure

train/
  patient0001/
    patient0001_2CH_half_sequence.hdf5
    patient0001_4CH_half_sequence.hdf5
  ...
val/
  patient0401/ ...
test/
  patient0451/ ...

Each HDF5 file follows the zea data format and contains:

• data/image/values — scan-converted B-mode sequence in dB, shape (n_frames, H, W), float32; x=0 at apex centre
• data/image/coordinates — per-pixel Cartesian positions in metres, shape (H, W, 3) [x, y=0, z]
• data/image_polar/values — polar-resampled B-mode sequence, shape (n_frames, n_r, n_theta), float32
• data/image_polar/coordinates — Cartesian [x, 0, z] positions of polar pixels in metres, shape (n_r, n_theta, 3)
• data/segmentation/values — multi-label bool segmentation, shape (n_frames, H, W, 4)
• data/segmentation/labels — ["unannotated", "LV_endo", "LV_myo", "LA"]; unannotated frames have only the first channel set
• data/segmentation/coordinates — same grid as image/coordinates
• metadata/subject — patient ID, sex, age
• metadata/credit — full citation string
• metadata/text_report — ejection fraction, frame rate, image quality
• metadata/annotations/view — "2CH" or "4CH" repeated for all frames
• metadata/annotations/label — "ED" / "ES" for the corresponding frames, "" otherwise
• metadata/annotations/image_quality — "Good" / "Medium" / "Poor"

License

CC BY-NC-SA 4.0 (https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode)

The CAMUS dataset is available free of charge strictly for non-commercial scientific research purposes only.

Citation

If you use this dataset, please cite:

@article{leclerc2019deep,
  title   = {Deep Learning for Segmentation Using an Open Large-Scale Dataset in
             2D Echocardiography},
  author  = {Leclerc, Sarah and Smistad, Erik and Pedrosa, Joao and Ostvik, Andreas and
             Cervenansky, Frederic and Espinosa, Florian and Espeland, Torvald and
             Berg, Erik Andreas Rye and Jodoin, Pierre-Marc and Grenier, Thomas and
             Lartizien, Carole and D'hooge, Jan and Lovstakken, Lasse and
             Bernard, Olivier},
  journal = {IEEE Transactions on Medical Imaging},
  volume  = {38},
  number  = {9},
  pages   = {2198--2210},
  year    = {2019},
  doi     = {10.1109/TMI.2019.2900516}
}

Links

• Original dataset: https://humanheart-project.creatis.insa-lyon.fr/database/#collection/6373703d73e9f0047faa1bc8
• zea toolkit: https://github.com/tue-bmd/zea