Synthetic imaging lab · no clinical connection

Imaging innovation, with a hard safety boundary.

This is the landing zone for future DICOMweb teaching workflows. Today it demonstrates a public-research teaching case and the architecture—without an NHS study, a patient identifier or a clinical inference endpoint.

Current stateSynthetic interface staged
DICOMweb
Not connected
Patient data
None
Live inference
Disabled
Teaching asset
Local

Future connection contract

Separate the public learning plane from any future clinical plane.

  1. 01
    DICOMweb teaching source

    Synthetic or explicitly licensed, de-identified public study only.

    OHIF boundary
  2. 02
    Browser interpretation layer

    OHIF and Cornerstone3D for 2D, MPR and structured teaching views.

    Not connected
  3. 03
    Research segmentation

    MONAI pipeline runs outside the public website with model, data and transformation provenance.

    Research only
  4. 04
    Human QA and release

    Orientation, overlay, transformation and use-boundary checks before a teaching asset is published.

    Required gate
  5. 05
    Local 3D and AR

    Reviewed GLB/USDZ teaching output with an independent static fallback.

    Demonstrated below

Public-research teaching case

A working browser 3D surface—not a clinical viewer.

Local 3D · public teaching asset

Move from a slice to a spatial model.

This orientation-checked ProstateView case demonstrates the reusable 3D layer inside the hub. The GLB, USDZ and fallback image are served locally; nothing here is diagnostic.

  • Drag to rotate and scroll or pinch to zoom.
  • Use the AR button on a supported HTTPS phone.
  • Open the full case to see provenance and QA status.
Open the full teaching case →
Axial public-research prostate MRI teaching slice with a segmentation overlay

Static teaching slice shown. Load the interactive model when ready.

Open building blocks

The software is ready before the clinical claim is.

Before any patient study

The next step is governance, not another demo.

  • Define intended purpose and explicitly exclude unsupported clinical uses.
  • Agree lawful basis, DPIA, retention, access control and de-identification validation.
  • Apply DCB0129/DCB0160 clinical-risk management where the system can influence care.
  • Complete model, transformation, orientation, accessibility and human-factors validation.
  • Obtain content, dataset and model-licence approval before distribution.