zeahub/cetus-miccai-2014

CETUS - 3-D Echocardiographic Ultrasound Dataset

This dataset is a zea-format (HDF5) conversion of the CETUS (MICCAI 2014) challenge data for endocardial segmentation in 3-D echocardiography.

Property	Value
Modality	3-D transthoracic echocardiography
Patients	45
Time points	End-diastole (ED) and end-systole (ES) per patient
Files	90 HDF5 volumes (45 patients x 2 time points)
Voxel spacing	Isotropic, ~0.576 mm (varies per patient)
Segmentation	Left-ventricle endocardial surface (binary)
Splits	train (1-30), val (31-38), test (39-45)

Conversion

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

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

Dataset structure

train/
  patient01/
    patient01_ED.hdf5
    patient01_ES.hdf5
  ...
val/
  patient31/ ...
test/
  patient39/ ...

Each HDF5 file follows the zea data format and contains:

• data/image/values - B-mode volume in dB, shape (1, depth, height, width)
• data/image/coordinates - voxel positions in metres, shape (depth, height, width, 3)
• data/segmentation/values - binary LV endocardium mask, shape (1, depth, height, width, 1)
• data/segmentation/labels - ["endocardium"]
• metadata/subject/id - patient name (e.g. patient01)
• metadata/annotations/label - time point: ["ED"] or ["ES"]

License

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

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

Citation

If you use this dataset, please cite the original CETUS paper:

@article{{bernard2016standardized,
  title   = {{Standardized Evaluation System for Left Ventricular Segmentation
              Algorithms in 3D Echocardiography}},
  author  = {{Bernard, Olivier and Bosch, Johan G. and Heyde, Brecht and
              Alessandrini, Martino and Barbosa, Daniel and Camarasu-Pop,
              Sorina and Cervenansky, Fr{{\'e}}d{{\'e}}ric and Valette,
              S{{\'e}}bastien and Mirea, Oana and Berber, Merih and others}},
  journal = {{IEEE Transactions on Medical Imaging}},
  volume  = {{35}},
  number  = {{4}},
  pages   = {{967--977}},
  year    = {{2016}},
  doi     = {{10.1109/tmi.2015.2503890}}
}}

Links

• Original challenge: https://www.creatis.insa-lyon.fr/Challenge/CETUS/
• Original dataset: https://humanheart-project.creatis.insa-lyon.fr/database/#collection/62eb991b73e9f0048c3a6c45
• zea toolkit: https://github.com/tue-bmd/zea