CAMELYON16
Cancer Metastases in Lymph Nodes Challenge 2016
About 400 H&E whole-slide images of sentinel lymph nodes from breast cancer patients at two Dutch hospitals, labelled for metastases with detailed tumour outlines, released under CC0 for the CAMELYON16 challenge.
Overview
CAMELYON16 is a collection of digitised lymph node sections from breast cancer surgery, created for the first challenge on detecting cancer metastases in whole-slide images. It was organised by Radboud University Medical Center and is widely used to develop and compare slide-level and pixel-level metastasis detection methods in computational pathology.
Composition
The slides are H&E-stained sections of sentinel lymph nodes. The data note on the combined CAMELYON dataset counts 399 slides for the CAMELYON16 part, while the challenge website speaks of 400:
| Center | Slides | No metastasis | Macro-metastasis | Micro-metastasis |
|---|---|---|---|---|
| Radboud UMC, Nijmegen | 249 | 150 | 48 | 51 |
| UMC Utrecht | 150 | 90 | 34 | 26 |
Slides containing only isolated tumour cells were not included in this part. The training data were released in two batches of 170 and 100 slides; a further 130 slides formed the challenge test set.
Acquisition
Cases were sampled from sentinel node procedures between 2006 and 2016, stratified so that metastasis-positive slides are over-represented. Radboud slides were scanned with a 3DHistech Pannoramic Flash II 250 and Utrecht slides with a Hamamatsu NanoZoomer-XR. Slides are stored as tiled, JPEG-compressed TIFF files at about 0.23 to 0.25 µm per pixel.
Annotations
A lab technician and a clinical PhD student outlined all metastases, and one of two expert breast pathologists checked every annotation. Consecutive sections stained for cytokeratin (immunohistochemistry) were used when the H&E slide was unclear. Outlines are provided as XML files and binary masks. The data note mentions that 15 slides may contain metastases that were not annotated.
Known limitations
- Only two hospitals and two scanners, all in the Netherlands.
- Positive slides are much more frequent than in clinical practice because of stratified sampling.
- Some slides contain two sections of the same node, of which only one is annotated.
- Patient demographics are not provided.
Cohort
Aggregate numbers from the sources below. Bars are relative to the largest value.
Condition
slides, values can overlap
- Lymph node metastasis 159 40%
Scanner vendor
slides
- 3DHISTECH 249 62%
- Hamamatsu 150 38%
Split
slides
- Training 270 68%
License and access
Our reading of the license, not legal advice. Before you use the data, read the original license and confirm that your use is allowed. We take no responsibility for how you use a dataset. Full disclaimer
Download without an account
Shared through GigaDB, the AWS Registry of Open Data and Baidu Pan together with CAMELYON17.
Creative Commons Zero 1.0 Universal
Public domain dedication. Do anything with the data, including commercial use, without asking and without having to give credit.
What you can do
- Yes
- Yes
- Yes
- Yes
What you can share
- Yes
- Yes
- Yes
What you must do
- No
- Share alike No
- No
- No
- Manuscript review No
- Release code No
- Return results No
- Delete after use No
Limits
- No
- Location limits No
Citation
Litjens G, Bandi P, Ehteshami Bejnordi B, et al. 1399 H&E-stained sentinel lymph node sections of breast cancer patients: the CAMELYON dataset. GigaScience 7, giy065 (2018). https://doi.org/10.1093/gigascience/giy065
All numbers
Every number on this page, as stored in stats.csv, with its source.
| Measure | Breakdown | Value | Source |
|---|---|---|---|
| Slides | total The challenge website states 400 whole-slide images | 399 | litjens2018 Data description section |
| Slides | vendor=3dhistech Radboud UMC slides (3DHistech Pannoramic Flash II 250) | 249 | litjens2018 Table 3; Data collection section |
| Slides | vendor=hamamatsu UMC Utrecht slides (Hamamatsu NanoZoomer-XR) | 150 | litjens2018 Table 3; Data collection section |
| Slides | condition=lymph_node_metastasis Macro-metastases 82 + micro-metastases 77 (both centers); no isolated-tumour-cell-only slides | 159 | litjens2018 Table 3 |
| Slides | vendor=3dhistech;condition=lymph_node_metastasis Radboud UMC: 48 macro + 51 micro | 99 | litjens2018 Table 3 |
| Slides | vendor=hamamatsu;condition=lymph_node_metastasis UMC Utrecht: 34 macro + 26 micro | 60 | litjens2018 Table 3 |
| Slides | split=train Two training releases: 170 (100 normal, 70 metastases) and 100 (60 normal, 40 metastases) | 270 | camelyon16-data-page |
| Slides | split=train;condition=lymph_node_metastasis 70 + 40 | 110 | camelyon16-data-page |
Sources
The keys used in the table above.
- litjens2018 Litjens et al. 2018, GigaScience (CAMELYON data note) paper
- camelyon16-data-page CAMELYON16 challenge data page website