MRI preview
Whole axial slice + gland / PZ / TZ / lesion overlay. Source: public Prostate158 dataset.
MRI preview is downsampled and not diagnostic. 3D AP/compass orientation is under QA; use PACS / the formal report for clinical review.
Demo guidetap to expand
- 1 Review the MRI preview tab first - establish where the lesion sits in 2D.
- 2 Open this 3D model tab.
- 3 Treat the 3D panel as experimental until AP/compass QA is resolved.
- 4 Toggle PZ / TZ / lesion layers to reveal each in turn.
- 5 Read the Lesion / context tab for volume and side notes.
- 6 Return to PACS and the radiology report for clinical review.
Layers
Toggle and dial opacity. Output is exploratory, not a diagnosis.
Orientation presets
Use preset views to translate radiology location into a 3D mental model. Camera-only - anatomy is unchanged.
Zone overlays
Anatomical orientation onlyZone overlays are for anatomical orientation only. They are not diagnostic segmentations unless explicitly labelled as source-derived.
- 20%
Translucent reference envelope.
- 50%
Where the majority of clinically significant lesions sit.
- 45%
BPH dominant. Anterior TZ lesions are easily missed on DRE.
- 100%
Source-derived for this case if hasLesion is true.
- n/a
Not available for this case. Not available for this case. CZ is rarely separated in routine prostate MRI.
- n/a
Not available for this case. Not available for this case.
- n/a
Not available for this case. Not available for this case.
- n/a
Not available for this case. Geometric 27-sector reference is currently shown only in the case summary, not the 3D view.
Report-to-model mapping
orientation onlyPaste a location phrase from the radiology report. ProstateView highlights side, zone, level and anterior/posterior - and nothing else. PI-RADS, staging and biopsy targeting are not assigned here.
Location mapping only. ProstateView does not assign PI-RADS, stage disease or replace the formal report.
Orientation rehearsal
Teaching and mental mapping only.
- 01 Read the locationRe-read the report sentence that describes the lesion. Note side, zone, level and anterior/posterior.
- 02 Anchor the sideConfirm which side is patient left and patient right on the orientation model.
- 03 Anchor the zoneToggle PZ vs TZ. Observe how the lesion sits relative to each.
- 04 Anchor the levelSwitch base / mid / apex. Match the level described in the report.
- 05 Rotate to the right viewUse a camera preset (posterior is usually most useful for PZ).
- 06 Look at the lesion in contextRead the case-summary card. Note the volumes and the sector string.
- 07 Return to PACSClose the orientation model and open PACS / the radiology report for the clinical decision.
This rehearsal is for teaching and mental mapping only. It is not biopsy targeting or procedural guidance. Return to PACS and the formal report for the clinical decision.
Case summary
- Source
- loading...
- Inference path
- -
- Structures
- -
- Lesion candidate
- -
- Confidence
- -
- Clinical use
- research and education only
Sector mapping is geometric, not predicted. PI-RADS is user-recorded against the radiologist's report.
Lesions
PI-RADS is captured from the radiologist's read, never inferred by the model. Lesion segmentation is low confidence (open-data Dice around 0.45).
Hand off to phone
/viewer/sample-public-prostate158-groundtruth-demo Open viewer here Scan from your phone to launch the AR-capable viewer.
Where this case came from
Public Prostate158 research dataset. See the global Method page for the full pipeline.
- Source: Prostate158 (Adams et al. 2022). Public research data only.
- Pipeline: outlines → marching cubes → decimation → GLB.
- Validation: technical comparison against public annotations, not clinical validation.
- See /method for the full plain-English explanation.
Safety and scope
Research and educational prototype. Not a medical device. Not for clinical decision-making.
Use alongside radiology report, MDT discussion, local protocols and senior clinical judgement.
- Public research datasets only. No NHS patient scans.
- PI-RADS is user-recorded against the radiologist's report. Never inferred.
- Lesion segmentation is exploratory and low confidence.
- Sector mapping is geometric, not predicted.
- No biopsy targeting or cancer detection claims. No treatment planning.
- MRI preview is downscaled and window-levelled for screen, not for clinical reading.
Detailed regulatory framing: docs/REGULATORY.md. Provenance: PUBLIC_DATASET_CASE.md.