Available Datasets

[ericspod]

Let's list the available datasets that we have access to.

• VerSe Dataset https://github.com/anjany/verse (labels are questionable)
• ADNI
• Decathlon

[Warvito]

• Chest x-ray - MIMIC-CXR Database and other datasets from physionet.org
• Brain images - UK Biobank
• Brain images - Most datasets from ida.loni.usc.edu (like PPMI, ADNIDOD, ABIDE, ...)
• Brain images - Human Connectome Projects and ABCD project from nda.nih.gov
• Several - RSNA's AI Challenge (in particular bone age challenge)
• Mammography - CSAW-M

Besides these, I tried to apply to RadImageNet, but no reply yet. Could anybody else also try to apply for it?

[danieltudosiu]

I think we need to make the difference between restricted and open-source datasets as I am not sure if generative model trained on restricted datasets (ADNI, UKB) can be freely shared for inference.

[marksgraham]

Besides these, I tried to apply to RadImageNet, but no reply yet. Could anybody else also try to apply for it?

I applied months ago and heard nothing

[danieltudosiu]

I have access to the following CT datasets in a curated version from Nvidia:

• Luna16
• TCIA COVID19
• HNSCC
• TCIA Colon
• LIDC

Furthermore, there are lots of datasets under the TCIA

[ericspod]

I would like to organise datasets that we have access to into three categories: those that can be accessed without any sort of permission or any other requests, those that do require requests but will give data out to anyone, those that are much more stringent in their requirements such as having clear scientific reasons or having a grant. I'm interested in the first because these, such as decathlon, we can use for public tutorials and information, the second category is useful for actual research as well. The third includes Biobank and others which will be the best ones to use but demonstrating anything with them isn't helpful to people without access.

[ericspod]

I have a few more that I know of, perhaps we should have a wiki page we can all add to as we find things. I know of many more sources for general machine learning datasets, such as tabular and time series data, but that's beyond imaging and now our focus quite yet.

Cardiac:

• http://www.cardiacatlas.org/
• ACDC: https://www.creatis.insa-lyon.fr/Challenge/acdc/databases.html
• M&Ms: https://www.ub.edu/mnms/

Cancer

• NCI CRDC Genomic https://datacommons.cancer.gov/data#key-datasets
• BRaTS:
  • https://www.med.upenn.edu/cbica/brats2020/data.html
  • http://braintumorsegmentation.org/
  • https://www.med.upenn.edu/cbica/brats/

Imaging

• Decathlon http://medicaldecathlon.com/

Alzheimer’s

• ADNI https://adni.loni.usc.edu/