Your data, your workflow¶
Choose how your tissue images reach Aurora. Processing is based in Switzerland, transfer is encrypted, and a local image workflow keeps original scans in your environment.
What a submission sends¶
patchesuploads the tile images, each with its position and scale.embeddingsuploads numbers computed on your own machine: a short row per tile, the tile's position, and the quality measurements that kept it.- Any covariate you add travels with the job, and every upload sends the file's name. Name it without identifiers.
The service stores what the submission needs: the observation, the covariates, the job's record and its result.
For how long¶
Once you submit an upload, it is permanent. No request in this API removes one, and nothing removes one on a timer. DELETE https://api.auroraomics.org/v1/jobs/{job_id} removes a job, its result and the job's own copy of what it read. The upload that job read stays. Decide before you submit.
| What | Kept for | Why |
|---|---|---|
| Upload never completed | a window the service reports | an upload still pending is removed once it has waited longer than that window; GET https://api.auroraomics.org/v1/ reports the window as retention.unsubmitted_upload_hours |
| Any other upload | no scheduled deletion | whether or not a job reads it; DELETE https://api.auroraomics.org/v1/jobs/{job_id} removes a job's own copy of what it read, immediately and audited |
| Result file | 90 days | long enough to fetch, re-fetch and archive; results are not stored indefinitely |
Download a result within its window, and keep it. Deletion does not remove copies already held in backups.
Landing a result in the workspace¶
land_in_workspace offers a finished result to the Aurora Intelligence inbox
of the same account, and landed on the job says whether it did. Importing it
into a workspace is your own step. That workspace's members then read it, and a
workspace you have made public is readable by any approved account.
How Aurora may use it¶
Under the academic terms, Aurora may use the data you upload, and the results generated from it, to develop, train, validate and improve its models and services, including those it offers commercially. A commercial agreement allows that only with your written permission.
Your data is never made public, and deleting it does not withdraw a model already developed with it.
Anonymisation is yours¶
Anonymise before you upload
Removing patient identifiers, personal data and protected health information is your responsibility, and it is a condition of use. That includes the metadata written inside a slide file. The service does not inspect images for identifiers and cannot remove them.
Cutting the tiles on your own machine leaves the slide label, the macro image and the metadata a scanner writes behind.
Where it runs¶
Uploaded images, generated results and account information are stored on infrastructure in Switzerland, where analyses also run. Aurora's administrators may access these data when needed to operate the service, provide support or investigate a problem.
What you send is encrypted in transit, through the edge network that also serves the website. The website's Privacy Policy is the full statement, including where that network may process a request.
Your images can stay on your machine¶
Opening, tiling, quality control and packing a slide run where the slide is, and upload nothing. The package sends no telemetry: no usage counter, no crash report, no update check.
On the Your slides stay with you route, DeepSpot-H runs on your machine too, and no image is uploaded. The numbers it sends are derived from the image, and partial reconstruction of image content from such numbers is a known research capability.