The paper says the data is public. Good. There are 400 slides, up to 3 GB each, and you only want to see one.
So you start a download, and wait. Then you find a viewer that reads the format. Then you find out it was the wrong slide.
Public data should be easy to look at. Paste the address into your browser instead. The viewer reads only the part on your screen, straight from the repository. You see the image in a few seconds, at full resolution, and nothing lands on your disk.
How it works
Big images are stored in tiles, at several resolutions. The Find Nuclei Viewer asks the server for the tiles you look at, and nothing more. Zoom in, and it asks for the finer ones. A whole-slide image of several gigabytes opens with a few megabytes.
Your browser talks to the repository directly. No server of ours sits in between, and nothing is uploaded or stored.
Near or far
Before the first picture, the viewer asks the repository a few small questions: how the file is laid out, where the tiles are. Each one travels there and back. So the first open depends on where the data lives, and where you are.
| Repository | Where the data lives | Kind of server |
|---|---|---|
| GDC (TCGA) | Chicago, USA | API in front of the files |
| CAMELYON16 on AWS Open Data | Oregon, USA | Cloud bucket |
| Cell Painting Gallery | Virginia, USA | Cloud bucket |
| Image Data Resource, BioImage Archive | Hinxton, UK (EMBL-EBI) | The repository’s own server |
| Zenodo | Geneva, Switzerland (CERN) | API in front of the files |
From Europe, an IDR image opens in about 3 seconds, a TCGA slide in about 7, and a CAMELYON16 slide from Oregon in 10 to 30. From the US, the American ones are quicker. Once the image is open, zoom and pan fetch only the tiles on screen. How fast those tiles arrive still depends on where the server is and what kind of server it is.
Repository by repository
TCGA on the Genomic Data Commons
The Cancer Genome Atlas left more than 30,000 open slides on the NCI Genomic Data Commons [1]. Find a slide on the GDC portal, copy its file ID, and open it as https://api.gdc.cancer.gov/data/<file id>.
Open a lung adenocarcinoma slide, TCGA-05-4245. The address from the file’s page on the GDC portal works too.
CAMELYON16 on AWS Open Data
400 sentinel lymph node slides from the CAMELYON16 challenge, the classic set for spotting breast cancer metastases [2]. They sit in a public bucket on AWS Open Data.
The tumour outlines drawn by the challenge’s pathologists sit beside the slides, as ASAP XML, and open with them. Open tumor_001 with its metastases outlined, 3.2 GB.
Image Data Resource
The IDR serves many of its studies as OME-Zarr [3]: cells, tissues, embryos and whole plates. Each image opens by its address, and Z-stacks open in 3D too.
Open a zebrafish lateral line primordium from study idr0079, then click 3D.
BioImage Archive
The BioImage Archive keeps files as the authors uploaded them [4]. A Zeiss CZI opens as it is, and so do the OME-Zarr images the archive converts.
- Open a confocal CZI with nuclei and lipid droplets, from S-BIAD4045.
- Open the DAPI image of a Xenium slide, 51,000 pixels wide, from S-BIAD2339.
Cell Painting Gallery
The JUMP consortium imaged more than 1,000 Cell Painting plates, and the Cell Painting Gallery holds them as the instruments wrote them [5]. Opera Phenix and Yokogawa CellVoyager plates open from their folder.
Open a 384-well CellVoyager plate. More in Yokogawa CellVoyager Plates in the Browser and Opera Phenix Plates in the Browser.
Zenodo
Zenodo serves files by address too. Use the link that ends in /content.
Open a CZI test set by Sebastian Rhode: a 96-well plate in 5 × 9 tiles [6].
At a glance
| Repository | What opens | Try it |
|---|---|---|
| GDC (TCGA) | Aperio SVS slides | Lung adenocarcinoma |
| AWS Open Data (CAMELYON16) | Whole-slide TIFF, with ASAP XML outlines | Lymph node with metastases |
| Image Data Resource | OME-Zarr, 2D, 3D, plates | Zebrafish |
| BioImage Archive | CZI, OME-Zarr | Confocal CZI |
| Cell Painting Gallery | Opera Phenix and Yokogawa plates | CellVoyager plate |
| Zenodo | Any file the viewer reads, by its /content link | CZI plate |
Every format the viewer opens is in Supported Formats.
When a link does not open
The server has to send parts of a file on request, and let your browser read them. Most repositories do. Some do not: the OpenSlide test slides, for one, only download. Then download the file and drop it on the viewer. It still opens in seconds, from your disk.
Running your own bucket or server? What the Server Needs lists the two settings.
Share what you see
Found the interesting part? Copy the link. It carries the image address, the position, the zoom and your channel settings, so your colleague lands on the same cells. Still nothing downloaded, on either side.
Next time a paper says the data is public, take it at its word. Open it.
References
[1] Grossman RL, Heath AP, Ferretti V, et al. Toward a Shared Vision for Cancer Genomic Data. New England Journal of Medicine 375, 1109-1112 (2016). doi.org/10.1056/NEJMp1607591
[2] 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(6), giy065 (2018). doi.org/10.1093/gigascience/giy065. CC0.
[3] Williams E, Moore J, Li SW, et al. The Image Data Resource: A Bioimage Data Integration and Publication Platform. Nature Methods 14, 775-781 (2017). doi.org/10.1038/nmeth.4326. Study idr0079: idr.openmicroscopy.org/study/idr0079, CC BY 4.0.
[4] Hartley M, Kleywegt GJ, Patwardhan A, et al. The BioImage Archive: Building a Home for Life-Sciences Microscopy Data. Journal of Molecular Biology 434(11), 167505 (2022). doi.org/10.1016/j.jmb.2022.167505. Studies S-BIAD4045 and S-BIAD2339, CC0.
[5] Weisbart E, Kumar A, Arevalo J, Carpenter AE, Cimini BA, Singh S. Cell Painting Gallery: an open resource for image-based profiling. Nature Methods 21, 1775-1777 (2024). doi.org/10.1038/s41592-024-02399-z
[6] Rhode S. CZI (Carl Zeiss Image) dataset with artificial test camera images with various dimension for testing libraries reading. Zenodo (2022). zenodo.org/records/7015307. CC BY 4.0.