Beyond Epithelial Thickness: Comprehensive Profiling of Corneal Epithelial Remodeling and Its Diagnostic Value Across Keratoconus Severity: The Noor Multimodal Corneal Imaging Study (NMCIS)

Background: Keratoconus causes characteristic spatial epithelial remodeling that may not be captured by conventional thickness measurements. Novel morphometric indices from MS-39 AS-OCT may better characterize these changes. Purpose: We aimed to evaluate novel MS-39 AS-OCT epithelial remodeling indices and compare their diagnostic accuracy with epithelial thickness and Pentacam tomography across the keratoconus spectrum, using an objective reference standard. Methods: This retrospective diagnostic accuracy study included 1319 eyes of 1286 patients with same-day Pentacam and MS-39 imaging at Noor Eye Hospital (2023–2025). Eyes were classified from the clinical records and Pentacam front-surface indices, without use of any epithelial data, as normal (n = 861), forme fruste keratoconus (FFKC; topographically normal fellow eye of very asymmetric ectasia; n = 40) or definite keratoconus (n = 418), graded mild, moderate or severe with the ABCD system. Generalized estimating equations, receiver operating characteristic (ROC) analysis with DeLong confidence intervals and cross-validated cut-offs, and bootstrap-validated logistic models were used. Results: The central epithelial remodeling index (CERI) and central epithelial thickness range (CETR) increased with severity (+0.137 and +6.13 µm per severity grade; both p < 0.001). For definite keratoconus, CERI (AUC 0.932; 95% CI 0.913–0.948) and CETR (0.930) outperformed central minimum epithelial thickness (0.839; p < 0.001) but not BAD-D (0.972). Their AUCs rose from 0.666–0.677 in mild to 0.993–0.995 in severe keratoconus, a gradient that partly reflects the disease spectrum. For FFKC, the best epithelial AUC was 0.668 (6 mm epithelial thickness range) versus 0.728 for BAD-D, and adding five epithelial indices to tomography did not improve validated discrimination (out-of-bag AUC gain +0.000 for definite keratoconus and +0.012, 95% interval −0.080 to +0.083, for FFKC). Conclusions: CERI and CETR quantify epithelial remodeling and discriminate established keratoconus well, but they add little validated discrimination to Pentacam tomography; in FFKC, their discrimination was modest and any added value remains uncertain. They may complement, but not replace, tomography and biomechanics.