EyeModelHub
Home Subspecialties Oculoplastics Model card
§Model Card

PeriorbitAI

Automated quantitative measurement of eyelid and periorbital landmarks (MRD1, MRD2, brow and canthal heights)

Van Brummen A. · 2021 · USA · doi:10.1016/j.ajo.2021.05.007
Open DOI
Measurement/Regression CNN External photo Internally validated
More attributes
Anatomy · OrbitAnatomy · Periorbital soft tissueConditions · Periorbital measurementSingle-centern = NR patients · 418 images images
§01Overview

Brief overview

PeriorbitAI uses a ResNet-50 encoder and U-Net decoder to segment eyelid, brow, and canthal landmarks from standard external facial photographs and derive linear measurements (MRD1, MRD2, brow heights). Across 418 prospectively validated images at a single US center, automated measurements achieved Dice 0.82–0.96 with intraclass correlation 0.90–0.95 against human graders and mean absolute error of roughly 0.5–2 mm. Pretrained weights and code are openly available on GitHub. The model has not been validated externally on multi-institutional cohorts and was not deployed clinically. It is the most code-transparent oculoplastic measurement model in the literature to date.

§02Performance
Headline performance

Dice coefficients 0.82–0.96; mean absolute error ~0.5–2 mm vs human graders; ICC 0.90–0.95.

0.82–0.96
Dice coefficients
0.5–2
mean absolute error
0.90–0.95.
ICC
Full metrics — as reported
Dice coefficients 0.82–0.96; mean absolute error ~0.5–2 mm vs human graders; ICC 0.90–0.95.
Calibration · Uncertainty
Not reported.
Bias · Fairness
Not reported.
§03Data & Validation
Study designProspective cohort
Center typeSingle-center
PatientsNot reported
Images / eyes418 images
Split80/20 + retrospective hold-out test set + prospective hold-out test set
Reporting frameworks None stated
Age rangeNot reported
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 BSD-3-Clause
§05Quality Review · APPRAISE-AI
Total score
54 /100 Moderate
ClinicalDataMethodRobust.Report.Repro.
Sub-domain scores
Clinical relevance 4/4 100%
Data quality 13/24 54%
Methodological conduct 10/20 50%
Robustness of results 3/20 15%
Reporting quality 10/12 83%
Reproducibility 14/20 70%
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_vanbrummen_2021
Full author listOwen, J. P., Spaide, T., Froines, C., Lu, R., Lacy, M., Blazes, M., Li, E., Lee, C. S., Lee, A. Y., & Zhang, M.
JournalAm J Ophthalmol
DOI10.1016/j.ajo.2021.05.007
ConditionsPeriorbital measurement
Architecture
Model internals
Architecture familyCNN
Architecture detailConvolutional neural network (ResNet-50 encoder and U-Net-style decoder)
Funding & conflicts
Disclosures
FundingNIH/NIA R01AG060942; NIH/NEI K23EY029246, Research to Prevent Blindness Career Development Award; Latham Vision Innovation Award; unrestricted grant from Research to Prevent Blindness
Conflicts of interestAaron Y. Lee reports support from the US Food and Drug Administration, grants from Santen, Regeneron, Carl Zeiss Meditec, and Novartis, personal fees from Genentech, Topcon, and Verana Health, outside of the submitted work
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