Rainbone at Hack the Valley 11: We Built a 3D Bone Stress Test and Made the AI Prove Every Number

Description and use

Rainbone is a prototype that uses 3D modeling and AI to assess bone strength, delivering visual and textual results verified against facts.

Problem

Clinicians lack a plain answer to how strong a patient’s bone is until a fall occurs.

Outcome

Rainbone delivers a fully automated, AI‑verified 3D bone stress analysis that provides clinicians with visual, quantitative, and plain‑language insights, improving early fracture risk assessment.

Workflow

• 1. Load a de‑identified femur CT and segmentation into 3D Slicer to export a closed surface.
• 2. Convert surface to tetrahedral mesh with TetGen (60,819 tetrahedra, 17,361 nodes).
• 3. Run linear elastic solve in SfePy (800 N load, fixed distal 15 mm, E=10 GPa, ν=0.3).
• 4. Render results in browser with vtk.js, color by von Mises stress or displacement.
• 5. Provide cut‑away view, cohort risk panel, virtual press test, and STL download of broken region.
• 6. For live scans, send synthetic CT to FastAPI backend, mesh, solve, and store runs in Tiger Cloud.
• 7. AI layer receives only pre‑computed facts, checks outputs against rules, and falls back to lighter models if rate‑limited.
• 8. Ask Rainbone chat drawer runs a LangGraph StateGraph with stages: triage, evidence, writer, critic, audit.
• 9. Red‑team testing with 32 hostile questions, 30 passed, 2 fixed.

Evidence

Rainbone is a research prototype built on public, de-identified research data. It is not a medical device, and it never says whether or when a fracture will happen.

Source

The femur pipeline runs offline: 1. HFValid CT and the authors' segmentation go into 3D Slicer, which exports a closed surface. 2. TetGen turns the surface into 60,819 tetrahedra and 17,361 nodes. 3. SfePy runs a linear elastic solve: 800 N on the femoral head, the distal 15 mm fixed, one assumed material (E = 10 GPa, ν = 0.3).

Source