Corneal Densitometry Maps for Diagnosis of Common Corneal Diseases: A Masked Four-Reader Pilot Study.

PURPOSE: To evaluate whether qualitative corneal optical densitometry (COD) maps alone allow diagnostic discrimination of common corneal diseases in a masked multireader setting.

METHODS

In this retrospective diagnostic accuracy study, 500 COD maps obtained with Scheimpflug imaging were retrospectively identified and evaluated by masked readers following a standardized, prospectively defined grading protocol. After excluding 50 training images, 450 maps [90 each of healthy controls, dry eye disease (DED), keratoconus, Fuchs endothelial corneal dystrophy (FECD), and epithelial basement membrane dystrophy (EBMD)] were graded by 4 masked observers with varying experience and assessed for sensitivity, specificity, positive predictive value, and negative predictive value using a one-versus-rest approach. Intergrader agreement was evaluated using Fleiss kappa.

RESULTS

Diagnostic accuracy ranged from 68.0% to 83.1% across graders, differing significantly between most grader pairs (McNemar tests) without a consistent experience-related pattern. Intergrader agreement was substantial (κ = 0.63). Sensitivity ranged from 47.8% for DED to 95.6% for FECD, whereas specificity remained high (86.7%-98.9%). Diagnostic performance was highest for FECD and EBMD; most misclassifications occurred between healthy controls and DED.

CONCLUSIONS

COD maps allow moderate to high diagnostic discrimination of common corneal diseases in this selected, advanced-stage case series, particularly in structurally defined conditions such as FECD and EBMD. As FECD was assessed using posterior rather than anterior maps, its performance is not directly comparable to the other entities and reflects a depth-matched, layer-specific analysis. Although densitometry does not replace clinical examination, these findings support its potential as a diagnostic adjunct, warranting validation.