I didn't receive the activation link. How do I proceed?

First, please check your spam folder. If the email still hasn't arrived, use the "Forgot password" feature on the login page to request a new activation link.

How can I reset my password if I forgot it?

Click "Forgot Password" on the login page. You will be prompted for your registered email address, and a reset link will be sent to you. Follow the instructions in that email to set a new password.

I can't create an account.

We allow our partners, universities, and various institutions to use our web platform. If you are unable to create an account, it is likely that the domain of your email address has not been approved yet. If you believe you are eligible and still cannot register, please contact us at jaime.barrancohernandez@hevs.ch so we can review it.

Which file formats can I upload?

You can select .nii, .nii.gz, .zip and .7z files. Archives may contain NIfTI images or DICOM folders — the content is extracted automatically, and DICOM series are converted to NIfTI before segmentation. Only human T1-weighted images are supported.

Can I submit DICOM images directly?

Not as individual files — DICOM has to be inside a .zip or .7z archive, since a DICOM series is a folder of slices rather than a single file. The platform detects the DICOM folder and converts it for you.

If you prefer to convert beforehand, dcm2niix is the standard tool:

dcm2niix -z y -o /path/to/output /path/to/dicom_folder

The equivalent in Python, using the same library this platform uses:

import dicom2nifti
dicom2nifti.convert_directory("dicom_folder", "output_folder", compression=True)

How many cases can I submit at once?

Up to 10 cases per submission. A single archive may contain several cases, and the counter on the upload list shows how many have been detected. Files that would push you past the limit are rejected with an explanation.

Can I use the platform in several tabs at the same time?

No — please use a single tab. All of your work is kept in one working folder on the server, tied to your account rather than to a browser tab. Starting a new upload in a second tab clears that folder, so the first tab's results are no longer available: its viewers will fail to open and its download will no longer work. A results viewer that is already open keeps displaying what it has loaded, which can make the first tab look healthy when its data is in fact gone. Finish and download one run before starting the next.

How long does a segmentation take, and can I close the page?

Your job is submitted to our HPC cluster, so the total time depends on how many cases you sent and on how busy the cluster is — typically a few minutes once the job starts running. Keep the tab open until it finishes. A run is limited to one hour, after which the request times out.

What happens if I reload or leave the segmentation page while a job is running?

Returning to the segmentation page cancels any job you currently have running on the cluster, and the same is true of pressing Reset or starting a new upload. This is deliberate — it prevents abandoned jobs from occupying the cluster. If you reload by accident, simply upload your files again and restart.

The segmentation failed. What should I do?

Your working folder is cleared when a run fails, so start again with a fresh upload. Because the whole submission runs as one cluster job, a single problematic image causes the entire batch to fail — if a retry fails too, try submitting the cases one at a time to identify which one is responsible. Check as well that the images really are human T1-weighted scans.

Note the error message shown on the page, and if you have an archive from an earlier run, look at its logs/ folder — app.log and the .err file from the cluster usually say what went wrong. Please send those logs, together with the approximate time of the failed attempt, to jaime.barrancohernandez@hevs.ch. A failed run produces no archive of its own, but we keep a copy of your run log on the server and can look it up from the time you give us.

What is inside the results archive?

A single .zip, with one folder per image, named after it. Inside each folder:

• seg.nii.gz — the segmentation mask of both eyes;
• img.nii.gz — the image the mask is defined on. Open the mask against this file, not against your upload;
• left_axial_length.png and right_axial_length.png, if you extracted biomarkers.

And alongside those folders, covering the whole run:

• biomarkers.csv, a summary table of every case, if you extracted biomarkers;
• the logs of your run and the details of the inference, under logs/;
• the license agreement.

The labels look flipped or misaligned on my image — is the segmentation wrong?

Almost certainly not, and there is a file in your archive that settles it: open the mask against _img.nii.gz. That is the exact volume the labels were computed on, it ships with every result, and they are correctly oriented on it in every viewer. If they line up there, your segmentation is fine.

Whether the mask also lines up on the file you uploaded depends on what you uploaded:

• A NIfTI — yes, always. Nothing in the pipeline alters your image's geometry, and _img.nii.gz is your own volume.
• A DICOM series — usually not. It has to be converted to NIfTI before it can be segmented, and the converter always writes LAS, while a standard axial series is stored LPS. The mask and your series then describe the same scan with one axis reversed. A series that happens to be LAS already is unaffected.

That is why the same mask can look correct in Mango, which places it by patient coordinates, and mirrored in ITK-SNAP's Load Segmentation, which pairs voxel with voxel. The two viewers are disagreeing about grids, not about your results.

Which image is in my results archive?

Every case folder holds img.nii.gz — the exact volume the masks were computed on. Load the masks against that file and they will line up in any viewer.

If you uploaded NIfTI, it is your own volume. If you uploaded a DICOM series, it is the NIfTI that series was converted to, because a series has to be converted before it can be segmented and the masks are defined on the conversion. Either way, keep your own copy of what you uploaded — the archive gives you the reference image, not your original files back.

Do the logs in my download contain other users' information?

No. Each run writes to its own log, and only that log is placed in your archive. The logs describe your own submission — the files processed, the cluster job, and any errors — and never another user's activity.

Where is my medical data stored, and is it exposed?

Processing takes place on secured infrastructure at the HES-SO in Sion (Valais). Your uploaded images are used only to produce your results and are not retained afterwards. The resulting segmentations and biomarkers are archived on our secured server and are used to monitor and improve A-eye's performance. Nothing is shared with third parties, and no data is publicly accessible.

Does the results viewer keep my images?

No. The viewer only displays your images while the page is open — it stores nothing in your browser, and neither your uploaded scan nor the resulting segmentations are kept once you reset the session, start a new upload, or leave the page. Download the archive before you go: it is the only copy you will have.

Am I allowed to upload patient data?

Only if you are authorised to do so. By uploading, you confirm that the transfer complies with the ethics approval and Data Transfer Agreements covering your data — you remain solely responsible for that compliance. Both confirmations must be ticked before an upload can start.

The pipeline finished, but there is a problem with the results. How do I report a bug?

Please write to jaime.barrancohernandez@hevs.ch. Including the logs/ folder from your results archive helps considerably, as it records exactly what happened during your run.