Eyelid Morphometry AI Platform (DeepLabV3+)
Automated computation of MRD1, MRD2, and eyelid lengths from clinical photographs
▸ More attributes
Brief overview
This study developed a DeepLab V3+-based neural network system for automated eyelid segmentation and quantitative measurement of eyelid morphology from standardized facial photographs. The model segmented sclera, cornea, light reflex, and eyelid structures to derive MRD1, MRD2, and upper/lower eyelid lengths while also detecting abnormal eyelid contour regions using intercanthal-line partitioning. Across 300 retrospectively collected subjects with normal eyelids, ptosis, or Graves’ orbitopathy from a single Korean center, automated measurements showed excellent agreement with manual grading for MRD1 and MRD2 (ICC 0.937–0.972). The system was internally validated only, without external testing, calibration analysis, or clinical deployment, though the authors proposed potential smartphone and portable-device integration for future clinical use.
ICC vs manual measurement: MRD1 = 0.97; MRD2 = 0.94; upper length = 0.80; lower length = 0.74; narrow Bland–Altman limits of agreement. Auto
ICC vs manual measurement: MRD1 = 0.97; MRD2 = 0.94; upper length = 0.80; lower length = 0.74; narrow Bland–Altman limits of agreement. Automated processing time = a few seconds per image.
| Study design | Diagnostic accuracy study |
|---|---|
| Center type | Single-center |
| Patients | 300 |
| Images / eyes | Not reported |
| Split | 60/20/20 |
| Reporting frameworks | None stated |
| Age range | Mean 50.5 |
|---|---|
| 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_nam_2024 |
|---|---|
| Full author list | Song, T., Lee, J & Kyu Lee, J. |
| Journal | Sci Rep |
| DOI | 10.1038/s41598-024-51838-6 |
| Conditions | Graves' orbitopathy, blepharoptosis |
Model internals ▸
| Architecture family | CNN |
|---|---|
| Architecture detail | DeepLab V3+–based segmentation of ocular structures and a calibration marker |
Disclosures ▸
| Funding | National Research Foundation of Korea (NRF) grant (MSIT) (NRF-2021R1A2C1011351); Institute of Information & Communications Technology Planning & Evaluation (IITP) (MSIT) (2021-0-01341, Artificial Intelligence Graduate School Program (Chung-Ang University)) |
|---|---|
| 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)