Lacrimal Obstruction AI Platform (AS-OCT CNN)
Screening model to differentiate patients with lacrimal duct obstruction (LDO) from normal controls using anterior segment optical coherence tomography (AS-OCT) images
▸ More attributes
Brief overview
This study developed a deep learning system using nine pretrained CNN architectures (including DenseNet, ResNet, VGG, Inception, and Xception) with an ensemble strategy to classify anterior segment OCT (AS-OCT) tear meniscus images for lacrimal duct obstruction (LDO) vs normal eyes. Using 230 images (117 LDO, 113 normal) with 5-fold cross-validation and extensive augmentation, the best ensemble achieved AUC 0.824 with sensitivity ~84–88% and specificity ~55–63% across subgroups. Grad-CAM (Score-CAM) highlighted the tear meniscus as the primary discriminative region, aligning with clinical assessment. This was a single-center retrospective diagnostic accuracy study with internal validation only; no external validation, code, weights, or clinical deployment were provided.
Best single model (DenseNet-169): AUC 0.78; sensitivity 64.6%; specificity 72.1%.
Best single model (DenseNet-169): AUC 0.78; sensitivity 64.6%; specificity 72.1%. Best ensemble (7-model): AUC 0.82; sensitivity 84.8%; specificity 58.8%.
| Study design | Diagnostic accuracy study |
|---|---|
| Center type | Single-center |
| Patients | 172 |
| Images / eyes | 230 images |
| Split | 5-fold cross-validation (no fixed percentage split reported) |
| Reporting frameworks | None stated |
| Age range | 38-67 |
|---|---|
| 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_imamura_2021 |
|---|---|
| Full author list | Tabuchi, H., Nagasato, D., Masumoto, H., Baba, H., Furukawa, H., & Maruoka, S. |
| Journal | Cornea |
| DOI | 10.1007/s00417-021-05078-3 |
| Conditions | Tear meniscus, lacrimal duct obstruction |
Model internals ▸
| Architecture family | CNN |
|---|---|
| Architecture detail | Multiple convolutional neural-network models |
Disclosures ▸
| Funding | NR |
|---|---|
| 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)