PatchCamelyon

Overview

PatchCamelyon (PCam, Veeling et al., 2018) contains 327,680 96×96 RGB patches extracted from the Camelyon16 lymph node whole-slide images (WSIs). A patch is labeled positive if its center 32×32 region contains metastatic tumor tissue. The official splits come from disjoint sets of slides:

Train: 262,144 patches
Validation: 32,768 patches (valid in the release)
Test: 32,768 patches

Data Structure

When accessing an example using ds[i], you will receive a dictionary with the following keys:

Key	Type	Description
`image`	`PIL.Image.Image`	96×96×3 RGB patch
`label`	int	`0` = `negative`, `1` = `positive` (tumor tissue in the center 32×32 region)
`wsi`	str	Source slide, e.g. `camelyon16_train_tumor_104` or `camelyon16_test_032`
`coord_x`	int	Horizontal position of the patch in the source slide
`coord_y`	int	Vertical position of the patch in the source slide
`tumor_patch`	bool	Whether the patch contains tumor tissue anywhere, including outside the center region

The release ships gzipped HDF5 files. While a split’s cache is built, its image file is decompressed next to the download (about 7.3 GB for train) and removed afterwards.

Usage Example

from stable_datasets.images import PatchCamelyon

ds_train = PatchCamelyon(split="train")
ds_test = PatchCamelyon(split="test")

sample = ds_train[0]
print(sample.keys())  # {"image", "label", "wsi", "coord_x", "coord_y", "tumor_patch"}

References

Project page: https://github.com/basveeling/pcam
Data (Zenodo): https://zenodo.org/records/2546921
Paper: https://arxiv.org/abs/1806.03962
License: CC0 1.0

Citation

@article{veeling2018rotation,
  title   = {Rotation Equivariant {CNNs} for Digital Pathology},
  author  = {Veeling, Bastiaan S and Linmans, Jasper and Winkens, Jim and Cohen, Taco and Welling, Max},
  journal = {arXiv preprint arXiv:1806.03962},
  year    = {2018}
}

@article{bejnordi2017diagnostic,
  title   = {Diagnostic Assessment of Deep Learning Algorithms for Detection of Lymph Node Metastases in Women With Breast Cancer},
  author  = {Ehteshami Bejnordi, Babak and others},
  journal = {JAMA},
  volume  = {318},
  number  = {22},
  pages   = {2199--2210},
  year    = {2017},
  doi     = {10.1001/jama.2017.14585}
}