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

iOS Blepharoptosis Detection AI (MobileNetV2)

Binary classification of blepharoptosis vs normal eyelids from near-field facial photographs

Tabuchi H. · 2022 · Japan · doi:10.1007/s00417-021-05475-8
Open DOI
Classification CNN External photo Internally validated
▸ More attributes
Anatomy · EyelidConditions · BlepharoptosisSingle-centern = 606 patients · 1276 images images
§01Overview

Brief overview

This study developed an iOS-based deep learning system for automated blepharoptosis detection from eyelid photographs captured with an iPad Mini 5. The model used a fully retrained MobileNetV2 convolutional neural network trained on 1,276 eyelid images from 714 subjects at a single Japanese centre. Using 5-fold cross-validation, the system achieved 83.0% sensitivity, 82.5% specificity, 82.8% accuracy, and an AUC of 0.900 for distinguishing blepharoptosis from normal eyelids. Score-CAM heatmaps suggested the network focused on clinically relevant eyelid features. No external validation, open-source code, pretrained weights, or prospective clinical deployment were reported, and the authors noted the model was only evaluated on Asian eyelids.

§02Performance
Headline performance

5-fold cross-validation: AUC 0.90; accuracy 0.83; sensitivity 0.83; specificity 0.83.

5-fold cross-validation: AUC 0.90
—
0.83
accuracy
0.83
sensitivity
Full metrics — as reported
5-fold cross-validation: AUC 0.90; accuracy 0.83; sensitivity 0.83; specificity 0.83.
Calibration · Uncertainty
Not reported.
Bias · Fairness
Not reported.
§03Data & Validation
Study designDiagnostic accuracy study
Center typeSingle-center
Patients606
Images / eyes1276 images
Split5-fold cross-validation (no fixed percentage split reported)
Reporting frameworks None stated
Age rangeMean 71.5
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 none
○Pretrained weights none
○Live demo none
License Not specified
§05Quality Review · APPRAISE-AI
Total score
35 /100 Low
ClinicalDataMethodRobust.Report.Repro.
Sub-domain scores
Clinical relevance 4/4 100%
Data quality 10/24 42%
Methodological conduct 0/20 0%
Robustness of results 4/20 20%
Reporting quality 12/12 100%
Reproducibility 5/20 25%
▸ 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_tabuchi_2022
Full author listNagasato, D., Masumoto, H., Tanabe, M., Ishitobi, N., Ochi, H., Shimizu, Y., & Kiuchi, Y.
JournalGraefes Arch Clin Exp Ophthalmol
DOI10.1007/s00417-021-05475-8
ConditionsBlepharoptosis
Architecture
Model internals
▸
Architecture familyCNN
Architecture detailMobileNetV2 CNN
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+BD18)
← Back to Oculoplastics