Related Experiment Video
Updated: Aug 25, 2026
Full-Field Optical Coherence Microscopy for Histology-Like Analysis of Stromal Features in Corneal Grafts
Published on: October 21, 2022
Interobserver Variability in the Identification of Keratoconus Suspects Among Teenagers and Young Adults: Comparison
Stella Georgiadou1,2, Athanasios Zisimopoulos2,3, Costas H Karabatsas1,2,4
1Optics & Optometry Division, Laboratory of Optical Metrology, Department of Biomedical Sciences, University of West Attica, Athens, Greece.
Purpose:
To evaluate interobserver agreement between 2 experienced corneal specialists in classifying eyes as normal (N), keratoconus-suspect (KCN-S), or keratoconic (KCN) using Scheimpflug tomography, and to assess the discriminatory performance of tomographic asymmetry indices.
Methods:
A total of 206 eyes from 103 participants (15-25 years) underwent Scheimpflug corneal tomography. Two experienced corneal specialists independently classified each eye based solely on four-map displays, without access to quantitative tomographic indices (BAD-D, ISV, IHD, IVA). Interobserver agreement was assessed using Cohen κ statistic. Discriminatory performance was evaluated by receiver operating characteristic (ROC) analysis and area under the curve (AUC), with observer classifications retrospectively compared with BAD-D, ISV, IHD, and IVA.
Results:
Interobserver agreement was moderate (κ = 0.40, P <0.01). Agreement between observer classifications and BAD-D was substantial for observer 1 (κ = 0.70, P <0.01) and fair for observer 2 (κ = 0.35, P <0.01). Among the evaluated indices, BAD-D demonstrated the highest discriminatory performance (AUC = 0.94 and 0.85 using observer 1 and observer 2 classifications, respectively). In expert consensus cases, BAD-D achieved an AUC of 0.97. A BAD-D threshold of 1.10 yielded 90% sensitivity and 90% specificity.
Conclusions:
Moderate interobserver agreement highlights the diagnostic uncertainty of identifying early keratoconus-suspect eyes and may contribute to variability in reported keratoconus prevalence. Among the evaluated tomographic indices, BAD-D showed the strongest agreement with expert classifications and the greatest discriminatory ability, supporting its use as an adjunctive tool for early screening in young individuals.
