AI-assisted fixation design · Region-aware generation

Bone geometry in.Anatomy-matched plate out.

Starting from an intact, corrected, or reconstructed 3D bone model, the system plans fixation sites, builds a region-aware plate topology, and generates editable geometry for different anatomical demands.

See the workflowTalk to the team
3D
bone geometry input
AUTO
site-to-plate workflow
STL
editable geometry output
Automated plate generation
Anatomy
InputHand / metacarpal model
01Hand anatomy

Interactive·Auto-playing workflow — switch cases or select any stage

01 / Design principle

The system generates the design. Experts remain in control.

The workflow accepts the bone state required by the case — intact anatomy, a planned correction, or a reconstruction — and transforms fixation intent into geometry-aware plate design.

Human

Define the target

Provide the target 3D anatomy and the fixation intent.

System

Plan and generate

Propose fixation sites, build a region-specific topology, and construct controllable plate geometry.

02 / Workflow

From 3D bone model to fixation plate

One continuous automated workflow, from case-specific anatomy and fixation intent to editable plate geometry.

03 / Capabilities

One engine, different plate families

The automated pipeline stays consistent while topology, profile, and hole patterns adapt to each anatomical region.

01

A fixation site is more than a point

Using the local surface normal as an axis, the system samples the intersection between a cylinder and the bone so the hole boundary conforms to local geometry.

Cylinder × surface
02

Surface-aware plate paths

Each path is projected onto the bone and smoothed, while abrupt height changes are detected as bridge regions to reduce surface penetration.

  • Closest-point projection
  • Arc-length resampling
  • Bridge detection
03

Anatomy-specific plate families

Compact linear plates for small bones and wide-head, long-shaft constructs for osteotomy fixation follow different topology rules — not one universal template.

  • Region-specific topology templates
  • Variable plate profiles
  • Case-specific hole patterns
04

Traceable project structure

Models, holes, groups, connection paths, plate geometry, and parameters remain in one design tree and can be exported together.

  • Multi-model management
  • Grouping and batch editing
  • STL + JSON export

Group 01

Illustrative plate controls

Ready
5 APERTURES5 VA LOCKINGAO-STYLE LINEAR

04 / Interactive parameters

Automation proposes. Experts shape the result.

Each anatomical region starts from its own plate logic. Engineering and clinical teams can refine the generated profile, bridge scale, hole layout, and countersinks before regenerating the final geometry.

Adjust the controls — the plate in the viewer at the top of the page updates live.

Request a walkthrough

05 / Technology

One pipeline, multiple fixation scenarios

Case-specific anatomy and regional topology define the design; deterministic geometry turns it into a controllable 3D result.

01

Input

3D bone model

Intact, corrected, reconstructed, or procedure-specific bone geometry enters the workflow.

02

Planning

Fixation sites

The system proposes candidate screw positions from geometry and regional constraints.

03

Topology

Plate structure

Surface-aware paths and region-specific template rules define the plate family.

04

Output

Editable plate geometry

Anatomy-matched geometry prepared for expert review, simulation, and export.

06 / Roadmap

Core generation pipeline built. Validation next.

The automated model-to-plate workflow and multi-region demonstrations are in place. The next milestone is structured evaluation in representative workflows.

Phase 01 · Complete

Geometry foundation

3D visualization, surface interaction, parametric plates, geometric checks, and traceable export.

Completed

Phase 02 · Complete

Anatomy-aware generation

Automated fixation planning, multiple plate topologies, and editable geometry generation.

Engineering system ready

Phase 03 · Next

Clinical validation

Representative case studies, expert evaluation, workflow integration, and evidence generation.

Next milestone

07 / Selected research

Selected research highlights

Research spanning point-cloud intelligence, sparse-view CBCT, and low-dose 3D bone reconstruction.

  1. 01
    FPConv: Learning Local Flattening for Point Convolution

    IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)·2020

    3D VisionPoint-Cloud UnderstandingScene Analysis
  2. 02
    Task-Aware Sampling Layer for Point-Wise Analysis

    IEEE Transactions on Visualization and Computer Graphics (TVCG)·2022

    3D PerceptionShape AnalysisIntelligent Sensing
  3. 03
    Learning Deep Intensity Field for Extremely Sparse-View CBCT Reconstruction

    International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI)·2023

    Medical ImagingLow-Dose CTOrthopaedic Imaging
  4. 04
    Learning 3D Gaussians for Extremely Sparse-View Cone-Beam CT Reconstruction

    International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI)·2024

    Medical ImagingLow-Dose CT3D Reconstruction
  5. 05
    C²RV: Cross-Regional and Cross-View Learning for Sparse-View CBCT Reconstruction

    IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)·2024

    Medical ImagingMulti-Anatomy ImagingLow-Dose CT
  6. 06
    Real-Time Reconstruction of 3D Bone Models via Very-Low-Dose Protocols

    npj Digital Medicine·2026

    OrthopaedicsSurgical PlanningLow-Dose Imaging
  7. 07
    DeepSparse: A Foundation Model for Sparse-View CBCT Reconstruction

    IEEE Transactions on Medical Imaging (TMI)·2026

    Medical Imaging AILow-Dose CTCross-Anatomy Reconstruction

Research prototype

From 3D bone geometry to anatomy-matched fixation.

One automated workflow, adaptable across anatomy, plate topology, and fixation intent. Structured validation is next.

Talk to the teamAll solutions