Mask a volume. See the target.
A synthetic lab for the source's masked 3D-patch idea. Change the representation controls, predict visible context, and inspect the exact geometry. This is not a scan or a clinical model.
WebGL is unavailable. The readable telemetry below still computes the patch system and challenge state.
Conceptual voxels--
Total patches--
Visible context--
Masked targets--
Make a prediction before applying. The hidden target is computed from the same patch grid.
CHALLENGE 1 OF 3 / PREDICT 50% CONTEXT
Why this matters
Visible patches provide context. Masked patch representations become targets for a self-supervised prediction step; the original volume is conceptual and synthetic.
What this is not
No patient data, clinical inference, report generation, triage, AUROC reproduction, or medical advice is present. The source's model metrics are not recreated by this geometry lab.
patches = (grid edge / patch edge)^3 | masked = round(patches x mask ratio) | visible context = visible / patches x 100