Documentation

Quick Start Guide

Create an account, run a task, and review the result.
1
Create a free account on the Sign-up page. No credit card is needed.
2
Verify your email using the 6-digit code or activation link in the verification email.
3
Choose a model on Submit Task. Upload a supported scan, or set the output parameters for synthetic image generation, which needs no input scan.
4
Review the cost and submit. Choose Submit to queue for an uploaded scan or Generate Volume for a synthetic image. The button includes the count when submitting several tasks or volumes. Follow progress in Queue.
5
Preview or download the output in Results. Image outputs open in the viewer; reports and supported analysis outputs open in Text Results.

You can optionally rate a completed result to contribute to the public model leaderboard.

Input Requirements

Check the selected model's modality, anatomy, and input requirements before submitting.
  • Format: NIfTI .nii.gz, a DICOM folder, or another supported research imaging format. The upload form shows the accepted formats and converts supported alternatives to NIfTI.
  • Maximum file size: shown in the upload form.
  • Synthetic generation: no input scan is required. Choose the output size and other parameters supported by the model.
  • Privacy: remove identifying information before upload, including visible text or identifying anatomy where relevant. Conversion and header cleanup do not replace your own checks.
A model or modality mismatch can produce unsuitable results. Review the model's intended use and limitations before submitting.

DICOM folders

How the automatic DICOM-to-NIfTI conversion behaves.

On Submit Task, choose Select DICOM folder. Converted series are added as NIfTI volumes to the file queue. Review the queued scans before submitting.

  • Browser conversion first. When supported, conversion runs on your device and uploads the converted volume and associated acquisition metadata. If unavailable or unsuccessful, the original DICOM files are uploaded for server-side conversion, where you select the required series.
  • Privacy checks still apply. Remove identifiers before upload. Conversion discards or clears many header fields, but does not remove identifying information visible in the image. Server-staged DICOM files are temporary, with a 30-minute expiry and automatic cleanup.
  • Series support. Classic single-frame and Enhanced multi-frame CT/MR are supported. The browser converter can decode JPEG-family formats; the server fallback accepts uncompressed or RLE data and reports unsupported compressed series.
  • Limits and geometry. Folder uploads are limited to 2 GB and 4,000 files. Server conversion checks spatial positions, orientation, and slice spacing, and reports inconsistent series. Review the converted scan's geometry before analysis.

Outputs and downloads

What you receive depends on the model.
The model details list its output format:
  • Segmentation: a .nii.gz label volume for overlaying predicted structures on the scan.
  • Synthetic image generation: generated .nii.gz volumes at the requested output dimensions. There is no input scan to compare against.
  • Reports and analysis: text (.txt) or vector (.npy) outputs, depending on the task.
Tools that can read NIfTI include:
ITK-SNAP 3D Slicer FIJI / ImageJ nibabel (Python) SimpleITK

Review and edit results

Review image outputs in axial, coronal, and sagittal views. Research use only — not a medical device.
  • Navigation: click or Tab to a pane, then scroll, ↑/↓, PgUp/PgDn, Home/End; 1/2/3 select panes; press ? in the viewer for the full shortcut list.
  • Crosshair: Shift+click any pane to jump all three planes to that voxel, with its coordinates and true intensity (HU for CT) read from the raw data.
  • Window/Level: one-click CT presets (soft tissue, lung, bone, brain) or free sliders.
  • Label legend: searchable for large label sets, with per-label colours, show/hide, and bulk All/None that respects the active filter.
  • Editing & revert: the correction toolkit lets you fix a segmentation in the browser. The first save preserves the original segmentation so you can revert while the files remain stored. An Edited badge identifies corrected results, and edited downloads are named …_output_edited.nii.gz.

For interactive CT and MR segmentation, see the Nalvera iSeg guide for Interactive Studio.

Study Design

Group scans by study and patient, track measurements over time, and keep related results together.
1
Create a study in Study Design. Assign commonly used models to make them easier to select when sending scans.
2
Add patients using pseudonymised codes or IDs. You can edit them, move them between studies, or delete them.
3
Add scans and enter their dates. Primary scan labels are CT, MRI, PET-CT, and Other. You can attach secondary PET, SPECT, or MR volumes for comparison and reuse stored scans without uploading them again.
4
Send to a suitable model. Use a scan, patient, or study's send action, or select scans in Submit Task → From Study Design. Review the estimated cost and any modality warning before confirming. Each scan becomes a separate task.
5
Track results in place. Each scan shows task status and links to completed outputs. Withdraw a queued task from its status badge to have reserved credits refunded. Tasks also appear in Queue and Results.

Model reference

Inspect task options, input and output formats, model pricing, and downloadable labels.
The gift icon identifies tasks that support their developers through the partnership program.
Loading model reference…

Browse models →  Open a model page for its intended use, limitations, and research references.

Provenance & Reproducibility

Download the recorded model version and execution details for your methods section.
The manifest includes the model version recorded at submission, plus container and execution details when recorded. Download it from the task's Provenance action on Results, or via GET /api/jobs/<id>/provenance. Example values: 
{
  "schema": "nalvera-job-provenance/1",
  "job_id": "AB12CD34",
  "model_id": "totalsegmentator_ct",
  "model_version": "2.5.0",
  "docker_image": "totalsegmentator_v250",
  "docker_image_digest": "sha256:…",
  "executed_by": "gpu001",
  "edited": false
}
Results use the version recorded for the task. Unrecorded fields, including those on older tasks, are null in the manifest. This record supports reproducibility; retain your input, settings, and downloaded output alongside it.

Task status

Follow task status in Queue and Results.
  • Queued validated and waiting for processing; credits are reserved unless the task is free. Your queue position appears in Queue.
  • Processing the model is running. A cancellation request is handled by the worker; stopping is not immediate.
  • Completed outputs and metrics appear on Results.
  • Failed the error is shown and reserved credits are refunded. Use Retry task while the input remains stored. A retry is billed again, using a free run when eligible.

Nalvera Privacy Vault

In development; not yet publicly available.

Privacy Vault is planned as a free desktop app for macOS, Windows, and Linux to prepare CT/MR scans locally. Planned features include pseudonymisation, optional face and embedded-text removal, and encrypted exports. See the Privacy Vault page for details.

The platform already accepts encrypted .nvlt containers on Submit Task, Study Design, and Studio. They are decrypted on the platform before validation. Privacy Vault will provide the desktop export workflow; regular .nii.gz uploads remain available.

Citing Nalvera

If the platform contributed to your research, please cite it.

Cite both the platform and the original model paper where available. Include the model version and run date; the task's Provenance record can help document the methods you used. 

@misc{nalvera,
  title        = {Nalvera.AI: a cloud platform for medical-imaging AI},
  author       = {{van der Heyden, Brent}},
  year         = {2026},
  howpublished = {\url{https://nalvera.ai}},
  note         = {Model: TotalSegmentator CT v2.5.0. Accessed YYYY-MM-DD}
}

Replace the note field with the model and version you actually used, and the date you ran it.

Staying up to date

Platform updates are posted on @nalvera_ai: AI models added to the catalog, new platform features, maintenance windows, and research notes.