Abnormal Lid Position Assessment AI (OpenFace)
Automated measurement of vertical palpebral aperture (VPA) from 2D facial photographs
▸ More attributes
Brief overview
OpenFace applies open-source facial landmark detection and constrained local neural field models to automated assessment of abnormal lid position from standard full-face photographs. In a feasibility study of 128 eyes from pre- and postoperative ptosis images, AI-derived vertical palpebral aperture measurements showed good agreement with clinician measurements (Pearson’s r = 0.87; 94.4% of observations within 2 SDs on Bland–Altman analysis). The system functioned across variable image quality without specialized imaging hardware and supports real-time webcam-based inference. Source code and downloadable pretrained models are publicly available through GitHub, although no dedicated clinical deployment or external multi-centre validation was reported. The platform represents one of the earliest openly accessible AI facial-analysis approaches applied to oculoplastic assessment.
Automated vs manual VPA/IPD: r = 0.87 (r² = 0.76); 94% of values within Bland–Altman limits of agreement. Robust to wide variation in image
Automated vs manual VPA/IPD: r = 0.87 (r² = 0.76); 94% of values within Bland–Altman limits of agreement. Robust to wide variation in image resolution.
| Study design | Retrospective cohort |
|---|---|
| Center type | Single-center |
| Patients | Not reported |
| Images / eyes | 128 eyes |
| Split | Not reported |
| Reporting frameworks | None stated |
| Age range | >18 |
|---|---|
| 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_thomas_2020 |
|---|---|
| Full author list | Gunasekera, C. D., Kang, S., & Baltrusaitis, T. |
| Journal | Plast Reconstr Surg Glob Open |
| DOI | 10.1097/GOX.0000000000003089 |
| Conditions | Ptosis, blepharoptosis |
Model internals ▸
| Architecture family | Classical ML |
|---|---|
| Architecture detail | OpenFace landmark detection with custom post-processing, producing scale-normalized VPA/IPD ratios for ptosis assessment |
Disclosures ▸
| Funding | NR |
|---|---|
| Conflicts of interest | Peter B. M. Thomas is supported by the National Institute for Health Research (NIHR) Moorfields Biomedical Research Centre. |
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)