Module 01 — R30 · R31 · R32 · R33
OPG Panoramic Acquisition & Preprocessing
Distortion correction · U-Net alveolar segmentation · 64×64px ROI extraction · No CBCT required
Panoramic OPG radiograph loaded
2616×1456 px · 85 µm/px · 8 bit
✓ 98% of clinics equipped
Preprocessing pipeline (R30)
1
Automatic modality detection
Waiting...
2
Geometric distortion correction
Waiting...
3
U-Net alveolar segmentation (R33)
Waiting...
4
Metal artifact removal
Waiting...
5
64×64 px ROI extraction (mROI · aROI · dROI)
Waiting...
Progress0%
Module 02 — R2 · R8 · R9 · R17 · R32
Multi-Branch CNN — Bone Structural Parameter Extraction
3 parallel convolutional branches · Embedding E [256D] · PSO ∈ [0,100] · No variogram (R9)
CNN extractor architecture (CNFE-OPG)
α₁ · 3×3
32 filters · ReLU · max-pool 2×2
Micro-trabecular architecture 50–250 µm
Micro-trabecular architecture 50–250 µm
α₂ · 5×5
32 filters · ReLU · avg-pool 2×2
Meso-trabecular 250–750 µm
Meso-trabecular 250–750 µm
α₃ · atrous d=3
32 filters · dilation=3 · ERF 13×13
Macro-architecture 750–2250 µm
Macro-architecture 750–2250 µm
Concaténation
96D → FC-512 (BN, drop 0.3) → FC-256 → E [256D]
PSO = wᵀ·E + b ∈ [0, 100]
PSO = wᵀ·E + b ∈ [0, 100]
Explicit exclusion (R9) — FTO Medimaps protection
✗ variogramme γ(δ) = (1/2N)×Σ[I(x)−I(x+δ)]²
✗ log-log slope β · ✗ sill · ✗ range · ✗ nugget
✗ variographic curve area
✗ log-log slope β · ✗ sill · ✗ range · ✗ nugget
✗ variographic curve area
Bone Structural Parameter computed
—
PSO ∈ [0, 100] · correlated to in vivo ISQ
0 — Dense50100 — Atrophic
Enriched radiomic vector φ [273D]:
E [256D]
DF box-count
db4 × 12 coeff.
5 first-order stats
Pearson correlation r (calibration set)
—
p-value (n=34 implants)
—
Detected modality
—
CPU inference time
—
Module 03 — R3 · R11 · R20
Classification Osseuse & Prédiction ISQ Préopératoire
CNN_CLASS 12 layers · Softmax 3 classes A/B/C · Predicted ISQ with 95% CI · Peri-implant µ-strain
A
ISQ ≥ 70
Dense Bone · D1–D2
—%
Lekholm-Zarb Type I–II
B
ISQ 45–69
Intermediate Bone · D3
—%
Lekholm-Zarb Type III
C
ISQ < 45
Spongy Bone · D4
—%
Lekholm-Zarb Type IV–V
Pre-operative ISQ Prediction
—
Pre-operative ISQ · 95% CI: —
Validated correlation r = 0.72 · p<0.001 · n=34
0.72
AGIS (OPG)
vs
0.68
Manual CBCT
Complementary Data
Peri-implant µ-strain
—
CNN_CLASS confidence score
—
Recommended implantation protocol
—
Failure risk (5 years)
—
TYPE IV SENSITIVITY (atrophic bone)
94%
AGIS vs 87% manual CBCT standard
Module 04 — R4 · R12 · R13 · R15
Generative Design & Digital Twin
Adaptive TPMS-Gyroid topology · Sigmoid porosity gradient · Eeq ≤ 1.5 × E_os · STL/3MF model
TPMS-Gyroid topology visualisation
Class B → TPMS-Gyroid
Unit G-surface
cos(x)sin(y)+cos(y)sin(z)+cos(z)sin(x)=t
Sigmoid porosity gradient P(z)
Apex z=0 · 50%z₀ = 0.42 (cortical front)Neck z=1 · 65%
Calculated parameters (Class B)
Selected topology
—
Graded porosity
—
Calculated Eeq modulus
—
E_os_local (estimated)
—
Constraint Eeq ≤ 1.5 × E_os
—
Target pore size
—
Inflection point z₀
—
Output format
—
GRADIENT FORMULA (R13)
P(z) = P_min + (P_max − P_min)
/ (1 + exp(−k(z − z₀)))
k ≥ 4 · z₀ = front cortical détecté
/ (1 + exp(−k(z − z₀)))
k ≥ 4 · z₀ = front cortical détecté
Module 05 — R5 · R14 · R23 · R24
ZTM14N EFEM Implant — SLM Monolithic Manufacturing
Ti-19Nb-14Zr at% · E ≤ 40 GPa · EFEM thermal deployment at 37°C · BIC 100% at 8 weeks
Customised implant — generated model
Spécifications matériau & process
L1 — ZTM14N Alloy (granted patent EP+US)
CompositionTi-19Nb-14Zr at%
Elastic modulus≤ 40 GPa
vs Ti-6Al-4V standard2.75× closer to bone
Phaseβ-metastable
L2 — SLM Process (granted patent EP+US)
ProcessPBF-LB (Laser Powder Bed Fusion)
Layer thickness20–40 µm
Atmosphere< 5 ppm O₂
EFEM — Active deployment
Mechanism—
Trigger temperature37°C (corps)
Deployment depth Δ—
FunctionalizationTiO₂ nanoporeux · HAp · RGD · BMP-2
AGIS Clinical Report · Complete Analysis
Results & Comparison vs Manual CBCT Standard
Demonstration of the clinical and economic superiority of the AGIS system using panoramic OPG
Predicted ISQ
36
Classe B · IC95% [30–42]
Analysis time
3.2 min
vs 15–20 min standard
Cost per session
<5 €
vs 150–300 € CBCT+expert
Type IV sensitivity
94 %
vs 87% manuel
CLINICAL RECOMMENDATION
▸ Class B confirmed — Os intermédiaire (D3)
▸ ISQ prédit : 36 ± 6 (IC 95%)
▸ Implant deployable EFEM recommandé
▸ Topologie TPMS-Gyroid · Eeq = 12 GPa
▸ Porosité 50→65% gradient sigmoïde k=4.2
▸ Protocole : delayed loading 8 weeks
▸ RFA follow-up post-insertion D7 · D30 · D90
▸ ISQ prédit : 36 ± 6 (IC 95%)
▸ Implant deployable EFEM recommandé
▸ Topologie TPMS-Gyroid · Eeq = 12 GPa
▸ Porosité 50→65% gradient sigmoïde k=4.2
▸ Protocole : delayed loading 8 weeks
▸ RFA follow-up post-insertion D7 · D30 · D90
AGIS vs Standard comparison (all metrics)
ISQ correlation
Duration (min)
Type IV sensitivity
Cost (€/case)
Claims activated in this demonstration
R1(a-e)R2R3(a-d)
R4(a-f)R5(a-d)R8 α₁α₂α₃
R9 ✗var.R11 CNN_CLASSR12 TPMS-G
R13 sigmoïdeR14 EFEMR15 µ-strain
R20 A/B/CR23 ZTM14NR24 fonct.
R30(a-f)R31 TPSR32 modal
R33 U-NetR34 intégration
Educational demonstration in accordance with patent ZTM-AGIS-2026-rev3 · ZTi-Technologies Holding © 2026 · IP Counsel: IP Trust, Paris
PSO, ISQ, TPMS parameters are simulated for illustrative purposes · Clinical validation ongoing: LSPM-CNRS UPR3407
r=0.72 (p<0.001) obtained on n=34 mandibular implants · independent ex vivo validation
PSO, ISQ, TPMS parameters are simulated for illustrative purposes · Clinical validation ongoing: LSPM-CNRS UPR3407
r=0.72 (p<0.001) obtained on n=34 mandibular implants · independent ex vivo validation