🧠 Cranial Implant Design

AI-powered pipeline for patient-specific cranial implant generation. From CT scan to custom implant in seconds — not weeks.

100
Skull Dataset (MedShapeNet)
4
Morphometric Clusters
<0.4mm
Mean Surface Distance
12s
Pipeline Runtime

📊 Skull Clustering & Statistical Shape Models

CLUSTERING

Cluster Visualization

PCA & t-SNE projections of 100 skulls into 4 morphometric clusters (mesocephalic, dolichocephalic, large, brachycephalic).

 CLUSTERING

Feature Distributions

Box plots of morphometric features (cephalic index, volume, surface area) per cluster.

 CLUSTERING

Elbow & Silhouette Analysis

Optimal k determination via elbow method and silhouette scores.

 CLUSTERING

PCA Variance Explained

Per-cluster principal component analysis — variance explained by shape modes.

 3D VIEWER

Mean Shape — Cluster 0

Mesocephalic (n=70). Average-sized skulls, CI=84.2.

 3D VIEWER

Mean Shape — Cluster 1

Dolichocephalic (n=3). Long, narrow skulls, CI=71.7.

 3D VIEWER

Mean Shape — Cluster 2

Large skulls (n=21). CI=78.0, Vol=41 cm³.

 3D VIEWER

Mean Shape — Cluster 3

Brachycephalic (n=6). Wide, short skulls, CI=93.5.

🔬 PoC v2 — Realistic Defects

V2

Overview — All Defects

Side-by-side comparison of craniectomy, trauma, and tumor defects with generated implants.

 V2

Decompressive Craniectomy

130mm frontotemporal flap, 101.8 cm³ removed. Surface fitting: Dice=1.0, MSD=1.3mm.

 V2

Traumatic Fracture

Impact crater + radiating fracture lines. Surface fitting: Dice=1.0, MSD=0.4mm.

 V2

Tumor Resection

Organic shape with safety margin, ~55mm. Surface fitting: Dice=1.0, MSD=0.4mm.

🧪 PoC v1 — Geometric Defects

V1

Frontal Defect (40mm)

Circular defect. Surface fitting: 6.4mm Hausdorff, 0.0% volume error.

 V1

Temporal Defect (50mm)

Irregular shape. Symmetry method: 4.1mm Hausdorff, best for this location.

 V1

Orbital Defect (30mm)

Eye socket region. Surface fitting: 6.0mm Hausdorff, 0.5% volume error.