Responsible use

The binding text is the published Terms and Conditions: https://auroraomics.org/terms. That page is the agreement and the only authoritative copy of it; what follows is a plain-language summary of the parts that decide how these models may be used, and it has no legal force of its own. Where the two differ, the published Terms are what you accepted. The Privacy Policy is at https://auroraomics.org/privacy.

Read the summary to know what you are agreeing to; read the Terms to know what you agreed to. They are kept to five points because those five are the ones that decide whether a use is allowed at all.

You must be allowed to upload it

You confirm you are permitted to upload the data, and that doing so breaks no institutional policy, ethical approval or third-party right. Getting the permissions is your responsibility, before the upload and not after it.

Anonymisation is yours to do

Data must be anonymised before it is submitted: no patient identifiers, no personal data, no protected health information — including metadata embedded in a slide file, which is the part people forget. Nothing here inspects your images for identifiers, and nothing here can remove them.

Not a medical device

This is a research platform. It is not cleared, approved or certified by any regulator, it does not comply with any medical-device framework, and it must not be used to diagnose, predict, treat or monitor anything, or to make any clinical decision. Non-clinical research use only.

The outputs can be wrong

A result is a model's guess. It can be wrong, it can be skewed by what the model was trained on, and it can be both while looking entirely reasonable — as can anything written about it, including this site. No warranty of any kind is given, and the risk of acting on a result is yours.

Cite the publication

Results are made available for non-commercial research on condition that you cite the publication named on the website's FAQ in any work that uses them. Not doing so can cost you access.

In short

A prediction from a stained image is a hypothesis about what an assay would have measured. It is useful for exactly the things a hypothesis is useful for: choosing where to look, screening a cohort, generating a question. It is not a measurement, it is not evidence about a patient, and it has no clinical use.

Whether a particular use counts as research use is a judgement only you and your institution can make. The answer to "may I use this to inform care" is already made, and it is no.