Frequently Asked QuestionsFAQ

Submitting a job takes about a minute once your account is verified:

1. Register on the home page — a Guest tier account is created automatically, no payment required.
2. Verify your email by entering the 6-digit code we send you.
3. Open Submit Job, drop in a NIfTI file (.nii.gz), a DICOM folder, or another supported format, pick an AI model, and click Submit to GPU Cluster.
4. Track progress on the Job Queue page — typical turnaround is under ten minutes.
5. When the job completes, the result appears in My Results with an in-browser image viewer and a download link.

If anything looks off, the documentation covers each step in more detail.

Two primary options: a compressed NIfTI file (.nii.gz) or a whole DICOM folder — the platform converts DICOM to NIfTI for you on upload.

Several other research imaging formats are supported too — converted to NIfTI right in your browser before upload, so only the de-identified result is sent (the platform never receives the original file): NRRD (.nrrd), MetaImage (.mha, or .mhd + .raw), Analyze 7.5 (.hdr + .img), MINC (.mnc), FreeSurfer (.mgh / .mgz), and HDF5 (.h5 / .hdf5).

You can also upload a .nvlt file — an encrypted scan export from Nalvera Privacy Vault that the platform decrypts automatically on upload. (Privacy Vault, the tool that produces these, is still in development.)

For NIfTI, use the compressed .nii.gz form (not a plain .nii), within the platform's per-file size limit.

Most jobs complete in under ten minutes. Actual processing time depends on two factors:

• Image volume — larger images (more megavoxels) take proportionally longer.
• AI model — more sophisticated models require additional compute time.

You can monitor the status of your job in real time on the Job Queue page.

Our email provider (Brevo) sends your verification code the moment you request it, so a delay is almost always on the receiving side — some institutional and corporate mail servers hold or filter automated messages for a few minutes before delivering them.

First, check your spam, junk, and promotions folders — automated emails often land there. If nothing arrives after a few minutes:

1. On the verification screen, use the "Resend code" link (it appears after a short countdown). Resend a couple of times and an on-screen help note appears with a direct contact email, so we can activate your account for you.
2. Make sure you registered with an email address you can actually access.
3. The email also contains a one-tap activation link — you can use it later instead of the code, even after closing the page.

Still stuck? Use the contact page and include the email address you registered with.

Study Design organises your imaging into studies → patients → scans instead of a flat list of uploads. You upload each scan once — tagging its modality (CT / MRI / PET-CT / Other) and scan date — and it is stored on the platform so it persists across sessions.

From there you can dispatch a single scan, a whole patient, or an entire cohort to any AI model in one action. It's ideal when you want to run the same model(s) across many patients and keep the results grouped.

There are two ways:

• From the study — right-click a scan, patient or study (or use the send icons). Models you've assigned to the study appear as one-click options. You'll see the estimated credit cost and a modality check before confirming.
• From Submit Job — open Submit Job → From Study Design, tick the scans you want (or Select all), choose a model, and send. Each scan becomes one job. Up to the platform's batch limit of jobs can be queued at once.

If a scan's modality doesn't match the model, it's flagged but can still be sent after a quick confirmation.

Yes. While a job is still queued, click the ✕ on its status badge in the study to withdraw it — the credits are refunded automatically. Completed jobs can be previewed or downloaded straight from the scan, and every job also appears in Job Queue and My Results.

Deleting a scan or patient never stops a job that's already running — it keeps processing and stays in My Results, it just loses its link back to the study.

Each job carries two cost components:

• Flat processing fee — a fixed amount per job, regardless of image size. It includes a small partner contribution: for models marked with the “Gives back” icon, part of this fee is routed to the model's original developers (see the partnership questions below).
• Volume-based model cost — scales with the megavoxel count of your image and the rate of the selected AI model.

The combined euro cost is divided by the platform's credit exchange rate to give the number of credits charged. Monthly credits are always drawn first; top-up credits cover any remainder.

A full interactive breakdown — including the exact rates and an example calculation — is available on the Pricing page.

A free job waives the entire cost of one job — any model, any image size. Every new account starts with 10 free jobs. One batch submission consumes one free job; in the Interactive Studio, one full session (unlimited clicks and structures, until you leave the viewer or save) counts as a single free job.

Free jobs are separate from credits: they belong to your personal account, are drawn automatically before any credits when you bill a job to "personal", and jobs directed to a lab budget never spend them. When a job runs free, its would-be credit cost is shown struck through with a "Free job" label.

Occasionally the platform may run a free-jobs period — announced with a banner — during which every job is free for everyone and neither your free jobs nor your credits are consumed.

No. Your monthly quota resets on a rolling 30-day cycle (the window starts when your membership period begins — at sign-up for the free Guest tier); unused monthly credits do not carry into the next period. They are a recurring quota, not a permanent balance — every tier, including the free Guest tier, has its own monthly quota.

Top-up credits are different: purchased separately, they never expire and stay until used.

Yes. If a job fails due to a processing error, any credits that were reserved for that job are automatically returned to your account. You do not need to request this manually.

Note that "failed" here means the model produced no result (a technical error stopped the run) — not a poor or unexpected result on your data; Nalvera is not responsible for the scientific quality of a model's output on your specific dataset, so completed jobs are not refundable on those grounds.

There are two ways to purchase top-up credits:

1. Click the credit balance box in the bottom-left of the navigation sidebar.
2. Click your email address at the bottom-left of the sidebar, then select Purchase Top-Up Credits.

In both cases, you can choose a preset amount or enter a custom value. The minimum purchase is from €5 for members and €10 for guest accounts.

Yes. Go to Settings → Billing and click Manage in Stripe. You can cancel your subscription there at any time.

Your current membership tier and its benefits remain fully active until the end of the paid period. No partial refunds are issued for unused time.

Members receive a discount on top-up credits as part of their membership benefit — the higher the tier, the greater the saving versus the standard rate paid by guest accounts.

This means that beyond the monthly credit quota, members also pay less when they need extra processing capacity. The exact discount per tier is shown on the Pricing page.

This icon marks models and tasks that are part of the Nalvera partnership program. Every time you run a job on one of these, a portion of the flat processing fee is automatically routed back to the model's original developers — the open-source teams, research groups, and partners whose work makes the platform possible.

You can hover over the icon anywhere on the platform (model catalog, Submit Job page, documentation) for a quick reminder of what it means. The exact split between platform fee and partner contribution is shown on the Pricing page under "How your job cost is calculated".

Two important details:

• The partnership program is orthogonal to the model's source type — an open-access tool (like TotalSegmentator) can still opt into the program on a non-exclusive basis, while models distributed exclusively via Nalvera are always in the program by definition.
• A single parent model can have different partners per task — for example, one task may give back to one research group and another task to a different lab. In that case the icon appears only on the specific tasks that route a contribution.

There is no extra charge to you for choosing a partnership-program model — the contribution comes out of the standard flat processing fee, not on top of it.

It depends on the type of credit:

• Monthly subscription credits are a recurring quota, so they are forfeited when the account closes.
• Top-up (purchased) credits can be handed off to another user via a one-time transfer code — if they qualify (see below). You can also simply choose to forfeit them.

When you delete your account, if your purchased balance is above your free welcome credits, you may mint a transfer code. A few rules apply, by design:

• Vesting: only credits older than 30 days can be transferred. Anything topped up within the last month stays locked and cannot be handed off.
• Processing fee: a 5% platform fee is deducted, so the recipient receives the vested amount minus 5%.
• One-time, shareable code valid for 31 days: the code is shown to you exactly once at deletion. It can be shared with another user (or a lab manager), who redeems it to receive the credits into their own balance. After 31 days the code expires and the remaining credits are gone.

These limits (vesting, the welcome-credit threshold, and the fee) exist so the transfer feature can't be abused to farm free starter credits across throwaway accounts.

Each upload must stay within the platform's per-file size limit. Files larger than this are rejected at submission.

If your image exceeds this limit, consider cropping the field of view or reducing the spatial resolution before upload. Make sure the result still meets the minimum voxel dimensions required by the selected model.

If you regularly work with files larger than the per-file limit, contact us — we can discuss options for your use case.

Yes. The batch upload on the Submit Job page accepts up to the platform's batch limit of image files in a single session. Each file is submitted as an independent job and processed separately.

Note that all files in a batch are processed with the same AI model and settings. If you need to apply different models to different scans, submit them as separate batches.

Upload speed is almost always limited by your local connection, not the platform. For the biggest improvement, connect via Ethernet with a good-quality cable instead of Wi-Fi, and turn off your VPN if possible — both have a major impact on speed. It also helps to pause cloud-sync or streaming apps and to restart your router if needed.

Keep in mind that home connections often upload much slower than they download, so contact your ISP if this is a recurring issue. To check your actual speed, use Speedtest.net — results are shown in Mbps; divide by 8 to get MB/s (e.g., 40 Mbps ≈ 5 MB/s).

Yes. Once a job completes, open My Results and click Preview to open the in-browser viewer — no download needed.

• Segmentation results — the predicted structures are displayed as a colour overlay on the original scan, with axial, coronal, and sagittal views. You can toggle individual label classes on and off and adjust the overlay opacity.
• Image generation results — the generated image is shown side-by-side with the input for direct comparison.

Image segmentation and adaptation tools are built into the platform so you can review and adjust your results before exporting. Adjust window/level, hide or merge label classes, and fine-tune outputs — all from within the viewer.

A Lab is a shared workspace for teams and organisations — research groups, companies, CROs, or any group of users who need a shared budget. A lab manager can invite members, top up a shared credit pool, and set per-member spending limits from a central dashboard.

Each member submits and owns their jobs individually under their personal account. The lab adds a shared financial layer on top — job data and results remain private per user.

If your organisation requires a formal Data Processing Agreement (DPA), we can arrange one on request. Contact us to discuss.

No. Your results are strictly private to you. Neither other lab members nor the lab administrator can access your jobs, input files, or segmentation outputs.

The lab admin can only see your aggregate credit spending within the lab budget — not the underlying data or any details about individual jobs.

Lab top-ups are billed at a flat per-credit rate for every lab, regardless of which manager initiates the checkout or what membership tier they hold personally. The exact rate is shown in the lab top-up dialog at checkout time.

The per-tier top-up discounts on the personal flow are an individual membership benefit and only apply when you buy credits for your personal wallet. They intentionally do not carry into the shared lab balance — otherwise a group could pool under a single high-tier manager just to extract a deeper discount on credits everyone consumes.

Access to uploaded data is restricted to you and a small number of authorized Nalvera staff, under access controls and logging. Other users, including members of your own institution or lab, cannot see your jobs or results.

• Uploads are stored on access-controlled, EU-hosted infrastructure; every transfer to and from the platform is TLS-encrypted.
• You can delete a job (and its inputs/outputs) at any time from My Results.

The complete data-handling rules are spelled out in the Privacy Policy.

No — and that is deliberate. For security reasons, Nalvera.AI does not offer a "remember me" or persistent (e.g. 30-day) login.

• Your sign-in cookie is a browser-session cookie — it is cleared as soon as your browser session ends.
• Even while your browser stays open, a session automatically expires after about a day, after which you simply sign in again.

This keeps the window small in which a forgotten, shared, or stolen session could be misused — which matters on a platform that handles medical imaging. Always use Sign out when you finish on a shared computer.

Yes — we strongly recommend fully stripping all header metadata from your imaging files before upload. For NIfTI, remove any field that could link the image back to an individual subject; DICOM folders are converted to NIfTI on upload and DICOM header attributes are discarded in the process.

The platform applies its own header sanitization as a secondary safeguard, but user-side anonymization is your first line of privacy protection. Standard tools such as dcm2niix and nibabel can help you inspect and clean NIfTI headers before submission.

• 90-day retention — datasets are stored for 90 days from upload and then removed automatically. This keeps Nalvera.AI affordable: medical-imaging files are large, orphaned datasets accumulate fast, and long-term storage would drive prices up for everyone.
• Extending storage — every dataset shows its deletion date in My Results and Study Design; extend any dataset for 1 credit per 5 extra days (monthly, top-up or lab credits). A warning email arrives 7 days before deletion. And as a side-benefit: your data does not linger on our servers.

Deletion is permanent and complete. The following are removed and cannot be recovered:

• The input file you uploaded
• The output file(s) generated by the AI model
• All associated job information — including job ID, selected model, processing duration, and submission date

Everything runs locally on your own computer. The app strips out identifying information from the scan's DICOM header — the patient's name, patient ID, birth date, referring physician, and hospital or institution name — while keeping technical details like the scan date so you can still track a patient's progress over time. This process is called pseudonymisation.

Nothing is uploaded automatically. You review the result and choose to export it yourself. See the Privacy Vault page for the full walkthrough.

A .nvlt file is what you get if you turn on encryption before exporting. Instead of saving a regular scan file, the app locks it so that only the Nalvera.AI platform can open it. Even if the file was intercepted or shared with the wrong person along the way, nobody without the platform's private key could read what's inside.

Encryption is optional — if you don't need it, export a normal scan file instead. When you do use it, .nvlt exports upload directly on Submit Job, Study Design, and the Interactive Studio; the platform decrypts them automatically before running your chosen AI model.

No, both are optional and switched off by default. Face removal can be applied to head CT and MR scans to strip out facial features so a 3D render of the scan couldn't be used to identify someone. Embedded text removal blanks out any text burned directly into the image, like a patient name or date stamp sometimes printed onto ultrasound frames.

You choose whether to turn either of these on before you export.

The app only ever holds a public key, which is like a lock that anyone can close but only Nalvera.AI has the matching key to open. Under the hood, this uses a well established public key encryption standard — the same type of cryptography that protects everyday tools like secure messaging apps and browser connections — combined with a strong modern cipher to lock the actual file contents.

When you encrypt a scan, the app uses that lock to seal the file. This means the app itself cannot open or read what it just encrypted, even if someone tried to force it to. Only the Nalvera.AI platform, which keeps the matching private key securely on its own servers, can unlock the file once it arrives.

This way, your data stays protected the whole time it's outside your computer, and there's no shared password or key sitting on your machine that could be lost, stolen, or misused.