1. 1

    Upload

    Drop a CT series in the browser, or send it DICOM-native straight from PACS, your contour station or the scanner via DICOMweb. No install, no TPS plugin, no vendor agreement.

  2. 2

    Queue

    Line up a single case or an entire day's caseload. Scans process back to back automatically — nobody needs to stay at the workstation to trigger each one.

  3. 3

    GPU inference

    Dedicated cloud GPUs spin up on demand and contour every slice — OARs and nodal levels — with zero user input. Head & neck OARs complete in under 3 minutes.

  4. 4

    Review, edit & approve

    A clinician checks every structure in Contour Station, correcting or adding where needed, then signs off structure by structure. Approval is a hard architectural gate, not a setting.

  5. 5

    RT Struct returns to TPS

    Approved contours are routed back to Eclipse, RayStation, Monaco or any DICOM-compatible TPS. Plan from them exactly as you would from hand-drawn structures.

Batch mode

Upload tonight. Review in the morning.

Queue the whole caseload before you leave. ContourOAI works through it patient by patient, so the team arrives to a full set of ready-to-review contours.

  1. 1

    8:00 PM — end of day

    Upload the batch — say 20 patients. Drag and drop the DICOM series or connect via DICOMweb; the queue starts immediately.

  2. 2

    8:00 PM → 9:00 PM — processing

    Each patient's OARs contour in under 3 minutes, back to back with no manual triggering. Twenty cases run in roughly an hour.

  3. 3

    9:00 AM — next morning

    All 20 patients are ready. The morning review dashboard shows batch status at a glance; the team reviews, approves and plans.

Timings measured under a stable, low-latency network connection. Actual processing time varies with network latency, image size and server load.

Integration

Pure DICOMweb, both directions.

What you send

A CT series, from wherever it already lives.

  • PACS — Orthanc, Sectra, Synapse
  • Directly from CT or MRI modalities
  • Browser upload for one-off cases
  • DICOMweb STOW-RS / WADO-RS

What comes back

An RT Struct, routed to the system you plan in.

  • Eclipse, RayStation, Monaco
  • Any DICOM-compatible TPS
  • Returned automatically on completion
  • No plugin and no version coupling

What stays put

Your existing infrastructure does not move or change.

  • PACS and TPS remain fully on-premises
  • No changes to clinical workflow
  • No new hardware for clinicians
  • Browser access from any computer or tablet

Practical questions

Running it day to day

Does our team need to install anything?

Only the on-premises DICOM gateway, which is a lightweight agent inside your network. Clinicians themselves install nothing — ContourOAI is fully browser-based and works from any computer or tablet.

How long does a case actually take?

Head & neck OARs complete in under 3 minutes, with other treatment sites at matching GPU-accelerated speed. In batch mode roughly 20 patients process in about an hour.

These timings assume a stable, low-latency network connection; actual processing time varies with network latency, image size and server load.

Can we correct the AI output, or is it take-it-or-leave-it?

Contour Station is built for correction. Clinicians edit slice by slice with brush and eraser tools, interpolate between edited slices, and add, split or merge structures — or contour any structure manually from scratch.

What if a structure we need isn't in the model?

Annotate your own cases and fine-tune a custom model on them. Any structure can be trained, including GTV and CTV tumour volumes, and the resulting model is selectable at contouring time alongside the base OAR models.