Residents versus experts in interpreting standard Scheimpflug imaging for the diagnosis of keratoconus

Purpose To evaluate the diagnostic performance of final-year ophthalmology residents in classifying keratoconus using standard Scheimpflug corneal tomography and to compare their accuracy to that of experienced corneal specialists. Design Prospective, investigator-initiated, multicenter observational study. Methods Five final-year residents assessed 1,239 anonymized Pentacam HR (Oculus GmbH, Germany) quad-map outputs, categorized as normal, suspect keratoconus, or keratoconus, previously obtained from 620 pediatric patients in a prior population-based imaging study. Two corneal experts independently established the reference standard. Residents rated their diagnostic confidence using Likert scales. Agreement was quantified using Cohen's kappa (κ) and Fleiss's κ. Results Resident-to-expert agreement ranged from Cohen's κ = 0.031 to 0.324 (slight to fair). Overall inter-rater reliability across all raters was slight, Fleiss's κ = 0.088. Four residents rated their diagnostic skill as neutral; one rated it as poor. Self-assessed competence relative to peers was rated as neutral by three residents and good by two. Conclusions Final-year ophthalmology residents showed limited agreement with expert graders when interpreting Scheimpflug images. The alignment between self-assessed confidence and measured diagnostic performance highlights a potential educational gap in image-based diagnosis of ectatic corneal disease. These findings support the need for competency-based curricular enhancements, including image interpretation benchmarks, expert-guided calibration sessions, and simulation-based learning modules, aiming to improve diagnostic accuracy and support earlier detection of keratoconus. Further multicenter studies are needed to validate these results and guide national and international training strategies.