EyeModelHub
Home› Subspecialties› Oculoplastics› Model card
§Model Card

Orbital Segmentation AI Model (2D U-Net)

Automatic segmentation of the orbital bony cavity on CT for fracture assessment and 3D surgical modeling

Morita D. · 2023 · Japan · doi:10.1016/j.jcms.2023.09.003
Open DOI
Segmentation CNN CT Internally validated
▸ More attributes
Anatomy · OrbitConditions · Orbital wall fractureSingle-centern = NR patients · 125 CT images images
§01Overview

Brief overview

This study developed a deep learning–based 2D U-Net model for automatic segmentation of orbital bones from CT images in patients with orbital fractures, enabling rapid generation of 3D printable anatomical models for surgical planning. Using 125 facial CT scans from a single Japanese institution, the model achieved a mean Dice coefficient of 0.860 and ASSD of 0.713 mm on internal validation cases. Four experienced surgeons judged most automatically generated models suitable for clinical surgical support with minimal modification. Code was released on GitHub under a modified non-commercial MIT-style license. The model was internally validated only, without external multicenter testing, regulatory clearance, or prospective clinical deployment.

§02Performance
Headline performance

Overall Dice = 0.86; ASSD = 0.71 mm.

Full metrics — as reported
Overall Dice = 0.86; ASSD = 0.71 mm.
Fractured side: Dice 0.85; ASSD 0.78 mm.
Healthy side: Dice 0.87; ASSD 0.63 mm.
Inference time ≈ 2 minutes for 10 cases.
Calibration · Uncertainty
Not reported.
Bias · Fairness
Not reported.
§03Data & Validation
Study designMethodological/development-only
Center typeSingle-center
PatientsNot reported
Images / eyes125 CT images
Split~92/8
Reporting frameworks None stated
Age range39-52
Sex (% female)Not reported
Race / ethnicity reportedN
Subgroup performanceN
§04Deployment
Deployment maturity
  1. Concept
  2. Prototype
  3. Internal val.
  4. External val.
  5. Regulator
  6. Clinical
Regulatory status
None No FDA / Health Canada / CE clearance.
Artifacts & license
◆Code repository ↗ ◆Pretrained weights ↗
○Live demo none
License MIT
§05Quality Review · APPRAISE-AI
Total score
45 /100 Moderate
ClinicalDataMethodRobust.Report.Repro.
Sub-domain scores
Clinical relevance 4/4 100%
Data quality 9/24 38%
Methodological conduct 4/20 20%
Robustness of results 7/20 35%
Reporting quality 12/12 100%
Reproducibility 9/20 45%
▸ Why this score?

The APPRAISE-AI tool grades AI/ML medical studies across 24 items in 6 subdomains, scored independently by two reviewers. The total reflects the methodological transparency of the publication — not the underlying clinical value of the model.

How to read this: high scores in clinical relevance and reporting with low scores in methodological conduct, robustness, or reproducibility is the dominant pattern in ophthalmic AI today.

Quality bands: Low <40 · Moderate 40–59 · High 60–79 · Very High 80+. Most published ophthalmic AI sits in the Low–Moderate band; studies that break into High typically combine multi-center datasets, prospective evaluation, and open-source release.

§06Full Metadata
Citation & identification
Bibliographic record
▸
Model IDoculoplastics_morita_2023
Full author listKawarazaki, A., Koimizu, J., Tsujiko, S., Soufi, M., Otake, Y., Sato, Y., & Numajiri, T.
JournalJ Craniomaxillofac Surg
DOI10.1016/j.jcms.2023.09.003
ConditionsOrbital wall fracture
Architecture
Model internals
▸
Architecture familyCNN
Architecture detailSeepened 2D U-Net
Funding & conflicts
Disclosures
▸
FundingNR
Conflicts of interestNo conflicts
Card lifecycle
Status, version, reviewers
▸
Card statusPublished
Card versionv1.0.0
First added2026-04-28
Last reviewed2026-04-28
Reviewersattribution opt-in pending
Update history
Card changelog
▸
  • 2026-04-28 initial publication (v1.0.0)
← Back to Oculoplastics