Signal Processing of Corneal Biomechanics in Early Keratoconus Detection: A Critical Review

Early keratoconus is difficult to detect because changes in the cornea may be subtle and diagnostic definitions differ between studies. Air-puff instruments such as the Ocular-Response Analyzer (ORA) and Corvis ST record the corneal response as waveforms or high-speed deformation images. Processing these recordings may reveal features not represented by conventional device indices. This critical review examines 36 primary journal studies of biomechanical measurement, signal processing, and keratoconus diagnosis. Evidence from established disease, subclinical and forme-fruste keratoconus, normal-appearing fellow eyes, and longitudinal progression studies is considered separately. Waveform-based and machine-learning methods have shown promising discrimination, particularly in established disease, but evidence for early detection is less consistent. Results were not pooled because the studies used different clinical definitions and validation methods. Measurement quality, device generation, intraocular pressure, corneal thickness and preprocessing can influence reported performance, while small cohorts and data leakage may produce overly optimistic estimates. Time-frequency analysis is technically feasible, but its additional clinical value has not been demonstrated. Future studies should define disease stages clearly, document processing methods, compare approaches in the same patient groups and evaluate fixed models in independent clinical settings. Clinical value will depend on more reliable detection of early disease and demonstrable improvements in patient care.