Home› Subspecialties› Oculoplastics› Model card

§Model Card

Orbital Fracture AI Platform (InceptionV3-XGBoost)

Automated binary detection of orbital blowout fractures on CT imaging

Li L. · 2020 · China · doi:10.1097/SCS.0000000000006042

Open DOI

Classification CNN CT Internally validated

▸ More attributes

Anatomy · OrbitConditions · Orbital blow out fractureSingle-centern = 188 patients · NR images

§01Overview

Brief overview

This study developed a two-stage AI system for detecting orbital blowout fractures on CT scans using a CNN–XGBoost pipeline. Inception V3 was trained on manually annotated transverse orbital CT slices to classify fracture vs non-fracture images, with preprocessing steps including HU conversion, bone windowing, and region cropping to reduce noise. Slice-level predictions were aggregated using XGBoost to generate patient-level diagnoses. On an internal dataset of 188 cases (single-centre retrospective cohort), the model achieved 92% slice-level accuracy and 87% patient-level accuracy (AUC 0.96). No external validation, calibration analysis, fairness assessment, or code/weights release was reported.

§02Performance

Headline performance

Slice-level accuracy = 0.92; patient-level accuracy = 0.87; patient-level AUC = 0.96.

Slice-level accuracy = 0.92

—

patient-level accuracy = 0.87

—

patient-level AUC = 0.96.

—

Full metrics — as reported

Slice-level accuracy = 0.92; patient-level accuracy = 0.87; patient-level AUC = 0.96.

Calibration · Uncertainty

Not reported.

Bias · Fairness

Not reported.

§03Data & Validation

Study design	Diagnostic accuracy study
Center type	Single-center
Patients	188
Images / eyes	Not reported
Split	80/20
Reporting frameworks	None stated

Age range	>/18
Sex (% female)	Not reported
Race / ethnicity reported	N
Subgroup performance	N

§04Deployment

Deployment maturity

Concept
Prototype
Internal val.
External val.
Regulator
Clinical

Regulatory status

None No FDA / Health Canada / CE clearance.

Artifacts & license

○Code repository none

○Pretrained weights none

○Live demo none

License Not specified

§05Quality Review · APPRAISE-AI

Total score

43 /100 Moderate

Sub-domain scores

Clinical relevance 4/4 100%

Data quality 14/24 58%

Methodological conduct 6/20 30%

Robustness of results 1/20 5%

Reporting quality 11/12 92%

Reproducibility 7/20 35%

▸ 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 ID	oculoplastics_li_2020
Full author list	Song, X., Guo, Y., Liu, Y., Sun, R., Zou, H., Zhou, H., & Fan, X.
Journal	J Craniofac Surg
DOI	10.1097/SCS.0000000000006042
Conditions	Orbital blow out fracture

Architecture

Model internals

▸

Architecture family	CNN
Architecture detail	Inception-V3 convolutional neural network with XGBoost aggregation

Funding & conflicts

Disclosures

▸

Funding	NR
Conflicts of interest	NR

Card lifecycle

Status, version, reviewers

▸

Card status	Published
Card version	v1.0.0
First added	2026-04-28
Last reviewed	2026-04-28
Reviewers	attribution opt-in pending

Update history

Card changelog

▸

2026-04-28 initial publication (v1.0.0)

← Back to Oculoplastics