Three-Dimensional Map of Descemet Membrane Endothelial Keratoplasty Detachment: Development and Application of a Deep Learning Model.
Source check: Source checked. Not independent scientific confirmation.
Finding: The model quantified graft detachment area and volume and was more sensitive than slit-lamp estimates in some substantial detachments; mean percent detachment exceeded specialists’ estimates by 8.2 percentage points (95% CI 6.2–10.2). Separate validation set: 20 scans/5,120 labeled images, Youden index 0.85, Dice 0.73 and R² 0.90 against manual labels. A prospective application set included 107 eyes; masked re-review of discrepant scans supported cases where clinicians underestimated detachment. No clinical outcome benefit was tested.
Technology: A supervised neural network segments posterior cornea and detached graft in anterior-segment OCT slices; the 3D maps aggregate the resulting cross-sectional segmentation.
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- A supervised neural network segments posterior cornea and detached graft in anterior-segment OCT slices; the 3D maps aggregate the resulting cross-sectional segmentation.
- Test and result
- The model quantified graft detachment area and volume and was more sensitive than slit-lamp estimates in some substantial detachments; mean percent detachment exceeded specialists’ estimates by 8.2 percentage points (95% CI 6.2–10.2).
- Limitation / still unknown
- Separate validation set: 20 scans/5,120 labeled images, Youden index 0.85, Dice 0.73 and R² 0.90 against manual labels. A prospective application set included 107 eyes; masked re-review of discrepant scans supported cases where clinicians underestimated detachment. No clinical outcome benefit was tested.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://doi.org/10.1016/j.xops.2021.100067
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 22 Nov 2026. Review scope: Read the primary full paper (publisher PDF copy via the University of Freiburg repository), including Methods and Results; checked both frozen catalogs for title/result duplicates. A U-Net segmentation model maps DMEK graft detachment from anterior-segment OCT, with separate manually labeled training and validation sets; the model was also applied to 107 prospectively followed eyes and identified detachments that clinicians underestimated. This is a specific clinical imaging result, not evidence of improved patient outcomes. The full text was accessed from an institutional repository copy after the PMC page presented a browser verification screen..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 3/5 · Validation 2/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.