Blepharoptosis Morphology Analysis AI (Attention R2U-Net)
Automated measurement of eyelid morphology (MRD1/2, lid lengths, corneal and fissure areas) from pre- and postoperative photographs
▸ More attributes
Brief overview
This study developed an Attention R2U-Net–based deep learning system to automatically segment eyelids and corneal limbus from standard facial photographs and quantify periocular morphology before and after blepharoptosis surgery. Trained on 4,138 eyes and tested on 135 ptotic eyes from a single Chinese center, the model achieved Dice coefficients of 0.962–0.964 and excellent agreement with surgeon-measured MRD1/MRD2 values (ICC 0.934–0.971), with repeat automated measurements showing near-perfect reproducibility. Beyond MRD measurements, the system quantified eyelid length, corneal area, and postoperative contour symmetry using MPLDs. No external validation, calibration analysis, regulatory approval, or public code/weights were reported, and the platform remained a research-stage image analysis tool.
Segmentation Dice: eyelid 0.96, cornea 0.96.
Segmentation Dice: eyelid 0.96, cornea 0.96. MRD agreement with manual: ICC 0.93–0.97; bias 0.09–0.15 mm. Excellent automated repeatability (ICC up to 0.999).
| Study design | Retrospective cohort |
|---|---|
| Center type | Single-center |
| Patients | 103 |
| Images / eyes | 135 eyes |
| Split | Not reported |
| Reporting frameworks | None stated |
| Age range | 6.3 |
|---|---|
| Sex (% female) | Not reported |
| Race / ethnicity reported | N |
| Subgroup performance | N |
- Concept
- Prototype
- Internal val.
- External val.
- Regulator
- Clinical
▸ 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.
Bibliographic record ▸
| Model ID | oculoplastics_lou_2021 |
|---|---|
| Full author list | Cao, J., Wang, Y., Gao, Z., Jin, K., Xu, Z., Zhang, Q., Huang, X., & Ye, J. |
| Journal | Ann Med |
| DOI | 10.1080/07853890.2021.2009127 |
| Conditions | Blepharoptosis |
Model internals ▸
| Architecture family | CNN |
|---|---|
| Architecture detail | Attention R2U-Net segmentation pipeline with automated landmark and scale detection |
Disclosures ▸
| Funding | National Natural Science Foundation of China (No. 82000948); National Natural Science Foundation of China (No. 81870635); Zhejiang Provincial Key Research and Development Plan (No. 2019C03020); National Key Research and Development Program of China (No. 2019YFC0118400) |
|---|---|
| Conflicts of interest | No conflicts |
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 |
Card changelog ▸
- 2026-04-28 initial publication (v1.0.0)